Skip to content

Latest commit

 

History

History
3070 lines (2286 loc) · 127 KB

RELEASE_NOTES.markdown

File metadata and controls

3070 lines (2286 loc) · 127 KB
Raw

Release Notes

2.8.1 (September 20 2017)

  • Check nodes number for durability checks. The store with durability requirements will report more specific error when the library cannot fulfill the condition during failover.

  • Handle enhanced error messages for subdoc operations. The subdoc responses will now expose context and reference ID if present.

  • Discover and bootstrap analytics service from cluster configuration.

  • Improve documentation of configuration parameters.

  • Enable Error Map feature by default.

  • Cleanup and extend minimal, libeventdirect, instancepool examples

  • Tools:

    • improve error reporting
    • experimental subcommand cbc-proxy
    • fix memory leaks
    • retry store operations during population phase in cbc-pillowfight

2.8.0 (August 31 2017)

  • Add support for OpenSSL-1.1.

  • Mask LOCKED status code for backward compatibility. This code (as well as others possible codes with 'item-locked' attribute) replaced with LCB_KEY_EEXISTS for SET, REPLACE and DELETE operations, and with LCB_ETMPFAIL for the rest.

  • Stop enumerating bootstrap nodes and mechanisms when the server returns authentication error.

  • Fixed double free error with lcb_ping3.

  • Exposed additional N1QL query parameters: lcb_n1p_readonly, lcb_n1p_scancap, lcb_n1p_pipelinecap.

  • Fixed cbc-subdoc/upsert without XATTR.

  • XERROR attributes synchronized with recent list on server.

  • Add missing documentation, and update stability of the API.

  • Do not throttle background configuration polling by throttle interval of configuration error handler.

  • Turn on background polling by default. The library will try to schedule configuration update every 2.5 seconds. To disable it use config_poll_interval=0.

2.7.7 (August 17 2017)

  • Implement new function lcb_ping3, which sends NOOP-like message to each service in the cluster and allows to measure latency along with health status of the connection. Might be useful for application-side keep-alive mechanisms.

  • Detect and expose bucket type through LCB_CNTL_BUCKETTYPE:

      lcb_BTYPE type;
  lcb_cntl(instance, LCB_CNTL_GET, LCB_CNTL_BUCKETTYPE, &type);

  • Fixed setting expiration in subdoc mutations.

  • Fixed DNS SRV support of Fedora 26 and FreeBSD.

  • Fixed DNS SRV with SSL connections.

  • Define EREMOTEIO in libuv

  • New subdocument command to remove whole document

  • New cbc command: cbc-subdoc. It provides interactive shell, where all subdocument commands accessible to inspect and modify documents in the cluster.

  • New cbc command: cbc-ping. It sends NOOP-like messages to all accessible services in the cluster, and displays the status along with latency.

  • Fix cbc-cat --replica, which now allows reading documents from replicas.

  • Implement NOOP command and cbc-pillowfight --noop, which sends NOOP instead of data manipulation commands.

  • Clarify errors found in .cbcrc. Now it will display configuration path along with error message.

  • Update examples:

    • Support username/password in subdoc and libeventdirect examples
    • Added example for subdoc XATTRs

  • Integrate fix for parallel build with dtrace on FreeBSD https://github.com/freebsd/freebsd-ports/commit/a71e1a86b851d42cd08319d9b28a4424e508e216

  • Make enhanced errors API public

  • Fixed various compiler and cppcheck warnings and documentation update.

2.7.6 (July 11 2017)

  • Expose enhanced errors for data commands. Couchbase Server 5 might return additional information about errors in the response body. According to SDK-RFC-28, the library allow user code to inspect this information using following functions:

    • lcb_resp_get_error_context(int, const lcb_RESPBASE *)
    • lcb_resp_get_error_ref(int, const lcb_RESPBASE *)

    They both return non-NULL strings if any of error information accessible. The lifetime of these fields limited by lifetime of the response object.

  • Report contextualized error messages during negotiation. The event reference could be used to find more details about authentication errors in the server logs.

  • Specify correct protocol level for SO_KEEPALIVE. This fixes setting tcp_keepalive option on connections.

  • Implement Error Map Retries. This implements the mechanics needed to retry commands on various errors based on dynamic settings supplied via the error map.

  • Add cluster admin provider. This provider doesn't do anything except serve as a source of management hostnames. And the library will fall back to it when bucket is not specified for cluster management connections.

  • Implement RBAC user management in cbc tools. In addition to examples/users, this can be a demonstration of new security APIs which appear in Couchbase Server 5.

  • Allow to inspect query errors in cbc-n1qlback. The command will write details for failed queries to file, specified with option --error-log.

  • Fix memory leak in io::Pool

  • Fix LCB_SDCMD_GET_FULLDOC. This would not actually work beforehand because the opcode it's mapped to is 0, and we used 0 as a sentinel value for an invalid opcode within the subdoc implementation.

  • Add LCB_NOT_AUTHORIZED error code. This error code maps to Memcached's EACCESS

  • Don't send empty Authorization header for HTTP requests, If there's no username and/or password

  • Internal refactoring:

    • io::Pool - remove empty dtor
    • Fix BadPluginEnvironment test on Fedora where libm.so is ld script
    • Add missing commands for cbc-help

  • Documentation update:

    • Add additional documentation for lcb_n1ql_cancel()
    • Typos

2.7.5 (May 17 2017)

  • Allow to disable sending the HELLO command when connecting to a server. Sending HELLO will cause a bootstrap failure with Couchbase Server 2.0 and older.

  • Fix error return value on reprepared query. Previously an error was returned if a N1QL query was reprepared, because the prior internal failure status was not updated.

  • Check for more N1QL error strings indicating the need to reprepare a statement.

  • Fix uninitialized memory issue when initializing lcb::Server and mc_PIPELINE

  • Couchbase 5.0 additions for Subdocument. This adds new protocol extensions for Couchbase 5.0. This consists of:

    • New LCB_SDCMD_SET_FULLDOC and LCB_SDCMD_GET_FULLDOC for full-doucument gets and sets via the subdoc API. This allows to access xattrs atomically with the document body.
    • New 'document flags'. These are in the form of LCB_CMDSUBDOC_F_.
    • Issues: CCBC-774

  • Fix bug where CCCP subsystem would be suspended indefinitely. CCCP subsystem would hang if an error was received for the config request itself.

  • Fix bootstrap with LCB_TYPE_CLUSTER. Previously bootstrap would fail because the client would not send proper credentials. Note that at this point, the default bucket must still exist.

  • Ignore empty DNS SRV replies. Some buggy DNS configurations return positive replies to SRV queries, but without actually containing any kind of A record as a response.

  • Enable background polling for configuration changes. This allows the client to periodically poll for configuration changes. This feature is disabled by default. You can use the config_poll_interval setting to enable it in the connection string.

  • Enable TCP Keepalive for newly created sockets. Newly created sockets have TCP keepalive enabled in order to avoid firewalls breaking connections due to inactivity. TCP Keepalive does not yet work for the libuv plugin (e.g. nodejs). You can use the tcp_keepalive=false directive in the connection string to disable it.

2.7.4 (April 18 2017)

  • Send SELECT_BUCKET command by default if server supports it. This enables new-style 'RBAC' authentication by default. In 2.7.3 users were required to use select_bucket=true in the connection string to enable this feature. In this version, the option is still available but is now mainly useful to disable it.

  • Improve lcb_AUTHENTICATOR API. This provides better documentation and some API/design/implementation fixes to the authenticator interface. The authenticator may be useful for integrators/wrappers who wish to correlate multiple credentials with their buckets. Note that the use of lcb_AUTHENTICATOR is not required for RBAC support. In order to use RBAC, simply make use of the username field in the lcb_create_st struct, or the username parameter in the connection string.

  • Fix bug where lcb_get_server_list() would return NULL.

  • Fix bug where client would not recover from failover. Clients from version 2.7.1 through 2.7.3 would not obtain a new cluster map after a node had been failed over (e.g. by hitting the "fail over" button in the UI).

2.7.3 (March 21 2017)

  • Provide the ability to send the SELECT_BUCKET when establishing a to a server. This is a building block allowing us to use 'RBAC'/username auth in the future. Note that this requires the select_bucket=true option in the connection string or equivalent, and that this feature as a whole is considered experimental.

  • Provide an option to disable DNS-SRV lookups. Because DNS SRV lookups often result in no result (i.e. NXDOMAIN) - which takes longer, allowing to disable such lookups may speed up startup time. This option is available via the connection string, using dnssrv=off

  • Send client/user-specific identifier in User-Agent HTTP header. The library already does this for data nodes (Memcached). Using it in HTTP services allows better supportability when diagnosing issues by reading the HTTP logs.

  • Fix bug where DNS SRV hostnames would not be used. While DNS SRV lookup was working, the library would not actually attempt bootstrap off those received hostnames.

  • Provide experimental Analytics support. This allows access to the Couchbase Analytics Service, available in some pre-release builds. API and syntax wise, Analytics is very similar to N1QL. To use the analytics service, set the LCB_CMDN1QL_F_CBASQUERY bit in lcb_CMDN1QL::cmdflags, and provide the appropriate host:port combination in the lcb_CMDN1QL::host field. - Currently, analytics support is not used in the cluster map/configuration.

2.7.2 (February 21 2017)

This release consists of additional internal refactoring and some improved logging messages. There is enhanced experimental XATTR support. This release also contains some bug fixes:

  • Fix build issues on FreeBSD. This allows normal BSD make to be used, rather than forcing gmake

  • Fixed broken JIRA link in README

  • Fix hanging SSL connections in IOCP/Completion mode. This would sometimes stall the connection by not requesting a write if a read was in progress. This would result in the command not being sent and the client hanging. Note that this only affects completion-style I/O plugins such as IOCP and libuv.

  • Rename LCB_SDSPEC_F_VIRTPATH to LCB_SDSPEC_F_MACROVALUES. VIRTPATH is intended for possible future materialized XATTRs.

  • Add LCB_SDSPEC_F_XATTR_DELETED_OK, which maps to the protocol flag of roughly the same name.

2.7.1 (January 19 2017)

This release consists of additional internal refactoring. More internals have been converted to C++.

  • Provide XATTR (Extended Attribute) prototype support. This provides a prototype implementation of xattrs, allowing the client to access extended (hidden) attributes of a document. This feature can be used on the client side by simply setting the LCB_SDSPEC_F_XATTRPATH bit in the lcb_SDSPEC::options field.

  • Add automatic DNS SRV record lookup when simple hostname supplied. The library will now automatically attempt to look up SRV records for various couchbase services if only one host is present in the connection string. Automatic lookup will not be performed if more than a single host is provded. See the Java Documentation on the matter (go to the bottom of the page).

2.7.0 (December 21 2016)

This release consists mainly of internal refactoring. Many of the internals have been 'upgraded' to C++

2.6.4 (November 28 2016)

  • Fix bug in pillowfight where large value sizes would cause a segfault.

  • Allow 64 bit values with cbc-incr and cbc-decr.

  • Fix encoding in lcb_n1p_setconsistent_token. This function would encode it as scan_vector but it should be scan_vectors.

  • Refactor negotiation internals to use C++. This is part of an internal refactoring to move our internals over to C++. This will make the code more manageable and extendable in the future.

2.6.3 (September 27 2016)

  • Fix memory corruption for some JSON APIs when no rows are returned. This fixes a bug where the JSON parser would read from garbage memory when parsing a response that had no rows, but due to a slow network, would be received in multiple chunks. This affects N1QL, CBFT, and View APIs.

  • Allow to adjust bytes to read per event loop iteration. This allows applications with high network throughput but low CPU capacity to prevent the library from oversaturating a specific event callback invocation or starve other sockets. It may be controlled through the read_chunk_size connection string option or via lcb_cntl_string.

  • Use htonll for CAS values. This allows a consistent representation of CAS values regardless of underlying platform. This allows interoperability between other SDKs with respect to exchanging CAS values. This however may break interoperability with older versions of the same SDK, if the CAS value is being passed around (which it shouldn't be).

  • New subdocument additions. This adds the LCB_SUBDOC_F_MKDOCUMENT flag which allows document creation if the document does not exist, and can be used for mutation operations which may create new paths or values. The LCB_SUBDOC_CMD_GET_COUNT is also added, which is a new command which retrieves the number of elements (for an array) or key-value items (within an object/dictionary) of a given path. Both these features require Couchbase Server 4.6 (or its prereleases).

2.6.2 (July 26 2016)

  • Don't crash on high number of FDs with select plugin. Because select(2) can only accomodate up to a certain number of file descriptors in the application, if opening a socket results in a too-high-numbered FD, the plugin will return an error rather than silently failing during polling.

  • Pillowfight can now set ttl (expiry). This is done via the -e or --expiry option.

  • Log URLs of HTTP requests. This may make it easier to debug some HTTP-based APIs. The URLs are printed as part of the TRACE logging level.

  • Fix crash on shutdown with completion-based I/O. The crash was a result of dereferencing the lcb_t after it had been destroyed. This bug affected completion-based I/O subsystems such as libuv and IOCP.

  • Do not require operation field to be set on lcb_CMDSTORE. Starting from this version, a new lcb_storage_t constant, LCB_UPSERT has been added with a value of 0. This means that upsert operations no longer need to explicitly use LCB_SET, it being the default.

2.6.1 (June 21 2016)

  • Index management API now properly handles 'fields' field. Previously this was treated as a csv string, when it is in fact a JSON array.

  • pillowfight now has a --populate-only option, which is useful when simply trying to populate buckets with large amounts of data.

  • Allow to bypass OpenSSL initialization. This allows applications which already have OpenSSL intialization code in place to suppress libcouchbase's own OpenSSL initialization code. You can disable SSL initialization by using ssl=no_global_init in the connection string.

  • Allow to toggle sending of multiple credentials in N1QL queries. You can use the LCB_CMD_F_MULTIAUTH in the lcb_CMDN1QL::cmdflags field to indicate that multiple credentials should be added. Otherwise only the current bucket's credentials will be sent.

  • Fix infinite loop on completion (UV,nodejs,IOCP) type IO plugins. This bug would be triggered when only a single server remained in the cluster and that single server failed. This would result in the client never being able to perform operations due to a delayed reference count decrement.

2.6.0 (May 17 2016)

  • Improve index management API and implementation. The rawjson field was being ignored and the condition field was missing as well.

  • Add pillowfight support for subdoc. At the simplest level you can simply invoke pillowfight as cbc-pillowfight --subdoc --json <other args>. Refer to the pillowfight documentation for more details.

2.5.8 (April 19 2016)

  • Fix SSL connectivity errors with views and HTTP bootstrapping. This would cause network connectivity issues when attempting to bootstrap via https:// or using views (lcb_view_query()). The fix is a workaround to not advertise older SSL encryption methods.

  • Do not abort when receiving a memcached EINVAL response. While the client should never end up in a situation where it receives an EINVAL from the server, it should nevertheless not terminate the execution of the current process. This bug was introduced in 2.5.6

  • Fix memory leak when using N1QL prepared statements. Prepared statements would never be freed, even when the client handle was destroyed (lcb_destroy()) causing a slight memory leak.

  • Append CRLF header after host header. This would sometimes result in odd HTTP headers being sent out.

  • Experimental CBFT (Full-Text search API) This version adds a new fulltext api. The API is a row-based API similar to N1QL and MapReduce. See the <libcouchbase/cbft.h> header for more details. The API is experimental and subject to change.

  • Allow additional client identifier for HELLO command. The SDK sends a version string to the server when doing initial negotiation. The server then uses this string in the context of any logging messages pertaining to that connection. In this version, a new setting has been added to allow 'user-defined' version strings to be appended to the logs. Note that this feature is intended only for use with wrapping SDKs such as Python, node.js and PHP so that their versions can be present in the log messages as well. This setting is exposed as a string control (in the connection string, or lcb_cntl_string() with the name of client_string.

  • vBucket retry logic changes. The client will now retry at constant 100ms rate when receiving not-my-vbucket error replies from the server (adjustable using retry_nmv_interval). It will also only use fast-forward map to determine the new location for the vbucket, and will not use extended hueristics.

    The most noteworthy user-visible change is the 100ms retry interval which will significantly decrease the network traffic used by the SDK during a rebalance.

    To restore the pre-2.5.8 behavior (i.e. use extended heuristics and and exponential retry rate), specify vb_noguess=false.

  • Add interface for multi-bucket authentication. A new API has been added to modify and add additional bucket/password pairs in the library. This is done using lcb_cntl and the LCB_CNTL_BUCKET_CRED setting.

    Note that this functionality is not yet used in N1QL queries due to MB-16964

2.5.7 (March 22 2016)

  • High-level index management operations. A volatile API for high level index management operations has been added to assist in common N1QL index operations such as creating the primary index and removing indexes.

  • Fix N1QL mutation token queries. This fixes some bugs in the previous implementation of the way mutation tokens were handled with lcb_N1QLPARAMS. The bugs and fixes only affect consumers of that API. Couchbase SDKs do not consume this API

  • Throttle config request retries on empty NMVB responses. This changes the previous behavior where a new configuration would be retrieved immediately upon a not-my-vbucket reply if a configuration was not included within the error reply itself. The new behavior is to request a delayed retry (i.e. subject to the default throttle settings) if the current configuration originated from the CCCP (Memcached) provider.

  • Rename LCB_CLIENT_ETMPFAIL to LCB_CLIENT_ENOCONF. This error code is returned only when there is no current client configuration. This error condition is not temporary and is actually fatal; a result of an initial bootstrapping failure. Note that the older name is still valid in older code for compatibility purposes.

  • Include PID in log messages on OS X. This makes the library logs (via LCB_LOGLEVEL etc.) easier to read on a mac. Previously this used to display only the thread ID, which was identical for multiple processes. Now the display reads as pid/tid, making it easier to read the logs in a multi-process environment.

2.5.6 (February 18 2016)

  • Sub-Document API (experimental) The client-side sub-document API has been implemented. Sub-document is a feature which vastly reduces network usage when operating on parts of documents. The API as it appears in this version is highly experimental and may (and likely will) change. Examples of use can be found in the examples/subdoc directory.

    • Priority: Major

  • Make lcb_sched_enter and lcb_sched_leave optional. When scheduling an operation (e.g. lcb_get3()), the scheduling function will implicitly create a scheduling context and submit the operation if none exists already. A scheduling context is explicitly created by calling lcb_sched_enter() and finished by calling lcb_sched_leave() or lcb_sched_fail().

  • API3 is now stable. The scheduling based API, introduced in version 2.4.0 and known as 'api3', is now stable and considered the API for use with the library. The previous API (i.e. 'api2') is considered deprecated.

    While API3 has been promoted to stable in this version, it has been available in its current form (and in a mostly compatible manner, except the implicit scheduling feature - CCBC-664) since 2.4.0.

    Storing an item in API2:

    lcb_get_store_t cmd = { 0 }, *cmd_p = &cmd; cmd.v.v0.key = "key"; cmd.v.v0.nkey = 3; cmd.v.v0.bytes = "value"; cmd.v.v0.nbytes = 5; cmd.v.v0.operation = LCB_SET; lcb_store(instance, NULL, 1, &cmd_p);

    Storing an item in API3:

    lcb_CMDSTORE cmd = { 0 }; LCB_CMD_SET_KEY(&cmd, "key", 3); LCB_CMD_SET_VALUE(&cmd, "value", 5); cmd.operation - LCB_SET; lcb_store3(instance, NULL, &cmd);

  • Add libcouchbase/ string to version identification to Memcached Connections to memcached will now be identified as libcouchbase/version rather than version. This increases readability for server logs

  • Hide mutation_token field from API3 mutation respones. The mutation_token field has never been part of the API itself (it was previously present when api3 was marked as "experimental"). The mutation token for any operation must now be retrieved using the lcb_resp_get_mutation_token() to retrieve the actual mutation token.

  • Server's PROTOCOL_BINARY_RESPONSE_EINTERNAL is no longer mapped to LCB_EINTERNAL. LCB_UNKNOWN_MEMCACHED_ERROR will be returned instead

  • Allow get-and-touch with an expiry of 0. Clearing a document's expiry with get is now possible, using the new LCB_CMDGET_F_CLEAREXP in lcb_CMDGET::cmdflags.

  • Allow multiple buckets when using sequence number consistency with N1QL This uses the new internal scan_vector protocol supporting multiple buckets, each providing their own lcb_MUTATION_TOKEN objects.

2.5.5 (January 12 2016)

  • Add retry_interval string option to adjust retry interval. This allows the setting to be modified via lcb_cntl_string() and specified in the connection string.

  • Handle backslashes in view row ID fields. This would previously not be handled correctly as the backslashes would not be removed, for example an ID of has_a_"quote would appear in the API as has_a_\"quote. This has been fixed and document IDs are now properly processed as JSON

  • Allow 'file-only' configuration mode. This allows applications to make the library instance exclusively configured from a file on the local filesystem rather than through network bootstrap. This feature is undocumented and unsupported. It may be enabled using the bootstrap_on=file_only connection string directive.

  • Log when squashing network errors. This will make the library log the original error whenever a network error is translated from a more detailed description into LCB_NETWORK_ERROR (in case detailed_errcodes is not enabled), or if an OS-level error is found which cannot be translated into a more specific library error.

    • Priority: Major

  • Fix memcached/ketama hashing This fixes a bug in the ketama hasing code which caused a key to be mapped to an effectively arbitrary server for the library instance. In practice the node a key was mapped to depended on the order in which the hosts were specified in the connection string. This has been fixed to always use hashing based on the lexical sort order of each server node. It is highly recommended that applications upgrade to this version (2.5.5) for proper memcached (cache) bucket functionality.

  • Add cbc-touch subcommand. This now allows the simple "touching", or modifying expiration time via the cbc command line client.

2.5.4 (November 25 2015)

  • Validate vBucket master nodes for bounds when receiving new configuration. This ensures that invalid configurations (addressing nodes which do not exist) do not make their way to KV routing operations.

  • Add lcb_strcbtype to print the name of the callback type This small convenience function is added to pretty-print the type of callback being invoked. The second argument to the callback can be passed to this function.

    • Priority: Minor

  • Disallow using certpath connection string option without explicit SSL (couchbases://) scheme. Since the SSL and non-SSL schemes are similar, a typo can let a user mistakenly think that SSL is being used. This is fixed by disallowing the other SSL option (certpath) when SSL is not enabled.

  • Add convenience function to retrieve hostname for key. This is an alternative to retrieving the vBucket configuration (via lcb_cntl()) and mapping the key to an index, and mapping the index to a node. Note that hostnames are sufficient for most but not all configurations. Those running multiple server instances on the same host (for example, cluster_run) will still need to use the full set of steps as this function does not return a port. This function is provided both as a vBucket API (lcbvb_get_hostname()) which retrieves the hostname for an index as well as a top-level instance (lcb_t) function (lcb_get_keynode()) which accepts a key buffer and length.

    • Priority: Minor

  • Ensure embedded jsoncpp does not throw exceptions. This caused some build issues with other build systems. This has been fixed by replacing any exception throwing code with abort() and assert() statements.

  • Log vBucket configuration parsing failures. This logs vBucket configuration parsing failures when a given config received could not be parsed. Parsing failures include both JSON syntax errors as well as improper fields or values within the config itself.

  • Allow per-request N1QL timeouts to exceed global timeouts. This scans the "timeout" property of the N1QL request and if set, will make the value of this property the timeout value for the request. A small parser was implemented which converted the N1QL timeout values (s, h, etc.) into microseconds.

  • Request configuration refresh when HTTP API redirects are received. Redirects in Couchbase Server are sent when a node is about to exit the cluster. We should take this as a sign to refresh the config since it indicates a node is about to be removed.

2.5.3 (August 27 2015)

  • Add N1QL timeout feature. This allows an independent timeout setting for N1QL. Previously this would use the views timeout.

  • Add N1QL prepared statements. This allows prepared statements to be used with N1QL. The library will maintain an internal "prepared statement cache" which contains cached responses for internal PREPARE requests. To use a prepared statement, an application can simply set the LCB_CMDN1QL_F_PREPCACHE bit in the cmdflags field within the lcb_CMDN1QL structure. All the rest is handled internally within the library.

2.5.2 (July 23 2015)

  • Fix off-by-one error when populating documents with pillowfight. Previously pillowfight would only populate N-1 documents where N is the (-I, --num-items) option. This has been fixed.

    • Priority: Minor

  • Don't generate negative keys for pillowfight. For certain option configurations, pillowfight would generate negative keys (some keys were in the format of -nnnnnn).

    • Priority: Minor

  • Allow in-progress N1QL requests to be cancelled. This allows in-progress N1QL requests to be cancelled by adding a new API, lcb_n1ql_cancel(). Invoking this function on an lcb_N1QLHANDLE handle (obtained via an out parameter in the command structure) will effectively stop the request and stop delivering callbacks to the user.

  • Rename lcb_SYNCTOKEN to lcb_MUTATION_TOKEN. This experimental (volatile) API has been renamed to "Mutation Token" to better reflect naming conventions found in other client libraries.

    • Priority: Minor

  • Implement histogram/timings information for N1QL queries via cbc-n1qlback. This adds support for the (-T, --timings) option in the cbc-n1qlback benchmark/stress-test utility. These timings reflect the latency between issuing the query and the receipt of the first row of the resultset.

  • Add (-M, --mode) option to cbc-create to allow for upsert, insert, etc. This allows cbc-create to use the full gamut of storage options available via the SDK by allowing an insert/upsert/replace mode as an argument to the new --mode option. append and prepend are now also provided as options, though not documented.

  • Support CBC_CONFIG environment variable for command line tools. This variable specifies a path to an alternate cbcrc file which may be used to provide cluster/bucket settings. This new option allows multiple configurations to coexist, without forcing any one of them to be inside the user's home directory.

2.5.1 (June 17 2015)

Bug fixes and improvements in 2.5.1

  • Fix hanging in durability operations if node is not present and constraints include failed node. This condition may be triggered when only a single node remains in the broadcast probe and a command sent to it could not be scheduled. A symptom of this bug was durability operations 'hanging'

  • Improved handling of topology changes when non-data (N1QL, Index) nodes are part of the cluster. This fixes some issues (mainly crashes) when non-data nodes are found inside the cluster during a topology change. While the library since version 2.4.8 was able to handle initial bootstrapping with non-data nodes, it would still crash when such nodes were encountered during configuration changes.

  • Improved random host selection algorithm for REST services This new algorithm ensures that the distribution is even among all eligible nodes for a given service. The old algorithm would only distribute evenly when the assumption that all nodes contained the same services were true. However this assumption is no longer necessarily true with Couchbase 4.0. In this case the algorithm ensures that the random selection inspects only the pool of nodes which are known to have a given service enabled.

  • Ensure ketama/Memcached-bucket hashing works correctly when non-data nodes are part of the cluster. In previous versions, ketama hashing would incorrectly consider all nodes as candidates for keys, which would result in some items being routed to non-data nodes, resulting in odd errors and inaccessible data. This is only an issue for the still-unreleased Couchbase 4.0.

  • Set TCP_NODELAY as a server side option, if it's enabled on the client. This uses the HELLO protocol functionality to enable this feature, if this feature is also enabled on the client (enabled by default).

New features in 2.5.1

  • Add cmake/configure option for enabling the embedding of the libevent plugin. This option, named --enable-embedded-libevent-plugin, will cause the plugin to be linked in with the core library (libcouchbase) rather than built as its own object

    • Priority: Minor

  • Add new combined "Store-with-durability" operation. This new API, called lcb_storedur3() allows specifying the storage input options as well as the associated durability options in a single command. Likewise, the status of the operation (including durability) is returned in the operation's callback.

2.5.0 (May 12 2015)

This change in the major version number signals the addition of new features for Couchbase Server 4.0; most of the actual new functionality for Couchbase 4.0 has already been included (incrementally) in prior 2.4.x versions. The new 2.5 version is built on the 2.4.x codebase.

  • Add cbc-n1qlback - a simple benchmark for N1QL queries. This functions by executing a line-delimited file containing query bodies using multiple threads if possible.

  • TCP_NODELAY functionality has now been placed into effect. This functionality was nominally present in prior versions, but would not work because of a typo.

    • Priority: Minor

  • Add 'tick' or 'pump' mode for I/O As an alternative to lcb_wait(), applications may call lcb_tick_nowait() to incrementally perform (non-blocking) I/O. This may provide a performance boost when batching/scheduling many operations. lcb_wait() itself must be called to guarantee completion of all operations, and the lcb_tick_nowait() functionality is only available on some I/O plugins. See the API docs for more information.

  • Allow "console logger" to log to a file As a convenience, it is now possible to direct the library to write to a log file rather than standard error. This is possible using the LCB_CNTL_CONLOGGER_FP (to programmatically set a FILE* value via lcb_cntl()) or console_log_file to set the path of the file (which will be overwritten) via lcb_cntl_string() or the connection string.

  • Make lcb_N1QLPARAMS repeatable/debuggable This allows the lcb_n1p_mkcmd() call to be invoked multiple times without actually modifying internal state. Previously calling this function twice would result in corruption of the internal parameter state. In this version, a new function, lcb_n1p_encode() has been added (which lcb_n1p_mkcmd() wraps) which may be used to inspect the encoded form of the query.

2.4.9 (April 14 2015)

  • Disable HTTP provider when any CCCP config is received. This makes the assumption that CCCP will always be available if even a single node provides an HTTP configuration. This change may break some corner-case upgrade scenarios from version 2.2 to 2.5 where a newly added 2.5 node is subsequently removed.

  • Fix additional missing defines for UV's EAI_* symbols This was not entirely fixed in 2.4.8, since some undefined macros still remained.

  • Make connection string timeout parameters (e.g. operation_timeout) always specify seconds; this will no longer require the decimal point to be used, but will break any prior usages of this value for microseconds.

  • Add cbc n1ql subcommand, which executes N1QL queries. This subcommand is still a bit rough around the edges, mainly because of server-side support for "pretty=false" (which makes the rows display rather weirdly).

  • Allow usage of -D option in cbc and cbc-pillowfight tools. This flag allows specifying connection string options in a more concise form on the commandline. The -D option may be specified multiple times in the form of -Doption=value.

    • Priority: Minor

  • Interpret couchbase://host:8091 connection string as couchbase://host Previously the library would treat 8091 as a memcached port. While technically correct according to the connection string semantics, would often be a source of confusion for users migrating from older versions of the library (or other SDKs) when using the form http://host:8091. A special provision is thus made for such a cas.

  • Implement enhanced durability using sequence numbers. This feature is available in Couchbase 4.0, and uses sequence numbers (optionally specified in the response packet of each mutation). sequence-based durability constraints help resolve some ambiguity in the case of checking the durability of items which have since been mutated, or in the case of a cluster failover. Using this functionality requires the LCB_CNTL_FETCH_SYNCTOKENS (or fetch_synctokens) and the LCB_CNTL_DURABILITY_SYNCTOKENS (or dur_synctokens) settings to be enabled (using lcb_cntl() or lcb_cntl_string(), or in the connection string). Enabling LCB_CNTL_FETCH_SYNCTOKENS will cause mutation response packets from the server to return an additional 16 bytes of sequence data, and enabling LCB_CNTL_DURABILITY_SYNCTOKENS will cause lcb_durability_poll() to transparently use this information (rather than the CAS) to check for persistence/replication. Only available in Couchbase 4.0. As a result of this feature, much of the durability subsystem itself has been rewritten, making durability overall more performant, even for CAS-based durability.

  • Add lcb_version_g extern symbol as alternative to lcb_get_version(). This symbol is an extern global which allows simple runtime checking of the library version. This is more convenient than lcb_get_version() as it avoids the requirement to create a temporary variable on the stack (lcb_get_version() returns a string, and requires an lcb_U32 pointer as its first argument to get the actual numeric version).

    • Priority: Minor

2.4.8 (Mar. 8 2015)

  • Retry next nodes on initial bootstrap, even if first node says bucket does not exist (or auth error), as this might be a recently removed node

  • The cbc and cbc-pillowfight binaries on Windows are now distributed in both release and debug variants. Previously they would be clobbered by one or the other depending on the build host. This fixes some issues in performance and dependency resolution when using these libraries.

  • Provide Read-Only config cache mode. In this mode the configuration cache file is read but never updated. Additionally, a missing file in this mode results in a hard error.

  • Keep vBucket heuristic guesses for limited periods of time. This will allow previously-learned vBucket master locations to persist over a configuration change, providing these changes were discovered recently. This allows the reduction of not-my-vbucket responses while allowing new configs to overwrite our heuristic info, if the heuristic is too old.

    • Priority: Major

  • Fix potential crashes in get-with-replica (lcb_rget3, lcb_get_replica) when there are no replicas available, or if there is an error in retrieving from one of the replicas.

  • Do not wait between not-my-vbucket retries This behavior restores the pre 2.4.0 behavior of retrying not-my-vbucket responses, with a more intelligent retry/rotation algorithm (see the release note about "vbucket map heuristics"). Previously a wait-time was introduced because of potential busy loops in retrying to the same node. The LCB_CNTL_RETRY_NMV_IMM setting can be used to disable this functionality (by disabling it, i.e. setting it to 0). This may also be disabled in the connection string via retry_nmv_imm=0.

  • Fix compilation error with UV when EAI_BADHINTS is not defined in the system. This is primarily an issue with newer UV versions and some versions of Linux

  • Allow means to disable C++ behavior on public library structures, allowing them to be initialized via C-style static initializers. This allows the zeroing of structures such as lcb_get_cmd_t cmd = { 0 }, which would ordinarily fail under C++ compilation because of that structure having a defined C++ constructor. Applications can take advantage of this feature by defining the LCB_NO_DEPR_CXX_CTORS preprocessor macro when compiling.

  • Fix some bugs in timing behavior (lcb_enable_timings). Timings between 1000-2000ms are now reported accurately. Additionally for more common handling, second timing ranges (between 1-9s) are reported in ms range (i.e. timings of 4 seconds are reported as 3000-4000ms ).

2.4.7 (Feb. 17 2015)

  • Fix SSL connection failures with SSL_UNDEFINED_CONST_FUNCTION. This would sometimes cause failures during early connection/negotiation stages.

  • Add experimental support for N1QL queries. This adds support for contacting N1QL endpoints and retrieving their result sets. The support at both the client and server components is still a work in progress. The API is similar to the view api (see lcb_view_query()) added in version 2.4.6. See details in <libcouchbase/n1ql.h>

  • Add experimental support for geospatial view queries. GeoSpatial views are available as an experimental feature in the current releases of the server. This will soon be offered as a stable feature in future releases. Applications may now use the lcb_RESPVIEWQUERY::geometry field and the LCB_CMDVIEWQUERY_F_SPATIAL to utilize geospatial views.

  • Fix memory leak for retried commands. In cases where a given command needs to be retried more than once, a memory leak was fixed in which the previous instance of the pacekt was not properly freed.

2.4.6 (January 20 2015)

  • Fix floating point exception on OS X. A floating point exception would sometimes be thrown on OS X sytems due to bad time structure initialization. The installation provided with homebrew for 2.4.5 fixed this issue. This is completely fixed in 2.4.6 Priority: Major

  • Improve warning messages when using deprecated options in cbc. This provides less ambiguous help messages when using deprecated options, showing a full and complete example for proper usage (when possible).

  • Add patch/micro version to DLL information on Windows. This lets the user see the exact version of the library on windows (via right clicking on the DLL and inspecting the details). Previously this information contained only the major and minor versions.

  • Provide pkgconfig (.pc) file with installation. This may help third party applications and libraries link against libcouchbase in some environments.

  • Provide one-off unsafe_optimize option for connection string/lcb_cntl. This provides a shorter way to enable some potentially unsafe optimizations which may make the client perform better in some scenarios.

    • Priority: Minor

  • Allow prompting for password in cbc. The cbc and cbc-pillowfight utilities will now securely prompt for the password if the password specified on the commandline is a hyphen (-).

  • Fix timeouts in some durability when not all replicas are online. The library will now fail the operation with LCB_DURABILITY_ETOOMANY rather than allowing the operation to timeout.

  • Add high level row-based view functionality. This adds a new API (currently considered volatile) which allows intelligently querying views. This builds atop the basic HTTP interface, and exposes a row-based callback API based upon streaming JSON parsing. The new API is defined in <libcouchbase/views.h>. This API will become more stable over time.

  • Parse configuration service locations for experimental services This exposes the N1QL and indexing services via the lcbvb API. See libcouchbase/vbucket.h for more information.

2.4.5 (December 17 2014)

  • Fix pillowfight ignoring set-ratio values above 50 The program would ignore these values and act as if 100 was specified, thus never issuing any GET operations

  • Building with autotools is no longer supported. If building the library from source, you must use CMake version 2.8.9 or greater. If unfamiliar with CMake, the README describes the process. Included also is a top-level script called configure.pl which functions with an autoconf-like interface.

    • Priority: Major

  • Fix customized IOPS crashes in some usage cases This fixes scenarios where applications assume that the built-in IOPS version is 0, and attempt to "Subclass" the IOPS structure. The internal version of the library structure is now 3, with some extra heuristics in place to ensure that the older code will still function. This issue was most visible in the Python SDK when using the gevent or Twisted plugins. This issue was first introduced with version 2.4.4

  • Allow raw certpath to be passed without need for percent-encoding (in most cases) This allows for a common pattern fo passing certpath in the connection string as a raw, unencoded path. This allows a user to do couchbases://host/bucket?certpath=/foo/bar/baz.

  • Fix missing installation UV plugin headers and source In 2.4.4 this was accidentally left out, and would only be installed if the plugin itself was built and installed. This affected building the Node.JS SDK using an existing libcouchbase install.

2.4.4 (Nov. 19 2014)

  • Detect disconnected pooled sockets This allows the connection pool to detect dead sockets which were closed by a server when they were idle. Sometimes servers will close connections to open idle sockets to save resources, or because of bugs in their implementations. This will fix some issues experienced with views where queries would randomly fail with LCB_NETWORK_ERROR or LCB_ESOCKSHUTDOWN, by first checking if the socket is alive before returning it back to the library's core. Note that the libuv plugin does not implement this functionality yet.

  • Fix pillowfight --min-size bug This fixes a bug where pillowfight would sometimes compare the min-size option to an uninitialized max-size option and round it down to that value; then would set the max-size option.

  • Don't ignore LCB_CNTL_DURABILITY_INTERVAL Fix a bug where this interval would be ignored, if modified by the user; always reverting to 100ms.

  • Fix memory leak with HTTP requests using a request body Requests (such as PUT, POST, etc) which contained a request body would cause a memory leak as the library forgot to free them when the request object was destroyed.

  • Fix errneous LCB_SUCCESS return when passed duplicate keys to lcb_durability_poll(). This would cause applications to mistakenly wait for a callback to arrive, when in fact the command had failed.

  • Add option to preserve vbucket ownership heuristics across config updates This allows the learned configuration settings to persist between configuration updates. The default behavior (up to, and including this change) is to discard any "learned" configuration in favor of the explicitly new config passed to the server. This new option allows this information to be persisted when a new configuration update is received. This behavior is considered experimental, and is primarily intended to reduce the time it takes for the client to relearn the current node (which is typically under 1-2 seconds).

  • Relocate memcached packets on topology changes for memcached buckets This enhances the behavior of the client when operating with a memcached bucket during a topology change. Previously the library would not relocate packets to new servers, resulting in errors for items which were now mapped to wrong nodes. The new behavior remaps the key to the new server using the updated ketama hashing. Note that as a current restriction, the remapping will be performed based on the key of the item, not any hashkey parameter being employed.

  • Return error if ignored/conflicting options are found This changes the behavior of the library to throw an error if a specific option field was filled in which did not make sense for a given command, for example, specifying a cas value using a LCB_ADD operation with lcb_store.

  • Fix issue when sending out large OBSERVE command. This would cause a partial command to be sent out if the size of the output packet was greater than 512 bytes. This has been fixed by dynamically growing the output buffer for OBSERVE

  • Fix spurious timeouts when using lcb_durability_poll This fixes an issue where sometimes the durability poll operation would prematurely time out.

2.4.3 (Oct. 21 2014)

  • Disable support for SSLv3 This works around the POODLE SSLv3 vulnerability by disabling support for anything below TLSv1.

  • Pillowfight enhancements Several behavior changes were made to pillowfight in this version. These are:

    • The -l or -c -1 option is in effect by default. This means that by default pillowfight will run an infinite number of cycles. The previous behavior was to default to a single cycle, requiring an explicit --loop to ensure the workload ran for a considerable amount of time.

    • When multiple threads are used, the workload is divided among the threads, thus making it that each thread only operates on a subset of the data.

    • A --sequential option has been added to allow the workload to operate in sequence on the total number of items. This is useful when wishing to load a bucket with many items.

    • A --start-at option has been added to allow the workload to specify an alternate range of keys; effectively allowing resumption of a previous run. The --start-at flag allows to specify the lower bound number which will be used to generate keys. Thus a --num-items=20000 and a --start-at=10000 will generate keys from 10000 through 30000.

    • The population phase has now been merged with the general workload implementation. This means that all worker threads will participate in the population phase. The previous behavior limited the populate phase to a single thread.

    • If stdout is detected to be a terminal, a simple "OPS/SEC" meter will periodically write the estimated throughput to the screen.

  • Fix memory leak when using large read buffers In the case where large read buffers are used (and the iovec elements becomes sizable, the library may end up incorrectly caching some memory blocks for future use. This fix makes the blocks be cached at the allocator level, so that they are properly (re) utilized.

  • Use forward map (and other heuristics) to get a next node for an item after a not-my-vbucket reply. Since the server (see bug attached) does not always guarantee that a given config is the most correct, the client must do some guesswork in order to properly map a node when it gets a not-my-vbucket; especially if the config says that the node is the correct one.

2.4.2 (Sep. 23 2014)

  • Mark the hashkey fields as being volatile. Usage of this field is not supported in many cluster systems and is thus not supported functionality. It exists primarily as a legacy from an older API

  • Add "key stats" mode to lcb_CMDDSTATS. This adds an additional key stats mode to the lcb_stats3() API which interprets the key field as being a document ID for which information (such as expiry, status) should be retrieved, rather than a system statistics key. Similar functionality already exists in the Java client library as getKeyStats(). In addition to this feature, a cbc stats --keystats option is also provided to employ this functionality from the command line.

  • Add more details about replica nodes in the cbc hash command.

  • Add lcb_cntl() setting to retrieve bucket name. Previously the library did not have a means by which the bucket name could be retrieved. Using the LCB_CNTL_BUCKETNAME setting, the bucket name will now be returned.

2.4.1

  • Implement mcflush subcommand for cbc. This was removed in the cbc rewrite as the previous flush command.

  • Requests issued to an invalid replica via lcb_get_replica() should fail with the LCB_NO_MATCHING_SERVER_CODE. Previously this sometimes went through due to an incorrect bounds checking in the lcbvb_vbreplica() function.

  • Fixed a memory leak in lcb_get_replica() when the operation would fail.

  • Fix memory leak in lcb_sched_fail() when extended commands are in the pipeline

  • Provide lcb_dump() function call to dump state information about a client handle. The function call itself is currently marked as volatile and the output format is very much likely to change.

  • Fix ratio argument in cbc-pillowfight. This ensures that the ratio argument will truly determine the ratio of gets to sets.

    • Priority: Minor

  • Fix crash when HTTP request is retried. This may take place during topology changes

  • Allow simple host-port string in connection string, giving it an implicit http:// scheme. This allows easier backwards compatibility with some application

  • Update some SSL options to better reflect server 3.0 functionality The old capath option has been renamed to certpath to indicate that the path is not to the signing authority, but to the self-signed server certificate generated by the server itself. Additionally the no_verify option has been hidden.

2.4.0 GA

  • [major] Attempt to retry items that are mapped to a non-existent node in a degraded cluster. Rather than returning LCB_NO_MATCHING_SERVER the behavior should be to wait for the item to succeed and attempt to fetch a new cluster configuration.

    In order to control how such 'orphaned' commands are handled, a new value has been added to the lcb_RETRYMODEOPTS called LCB_RETRY_ON_MISSINGNODE which dictates how commands should be rescheduled if the associated vbucket has no master. The default is to retry the command until it times out, but by setting this value to 0 (See LCB_CNTL_RETRYMODE) it may only be attempted once, causing 'fail fast' behavior in such a case.

  • [major] Don't throttle config requests based on initial file-based config. This allows the client to quickly recover from a stale config cache without waiting for the LCB_CNTL_CONFDELAY_THRESH interval to elapse. Prior to this fix, a client would appear to "not recover" if bootstrapping from a stale cache. In reality the client would eventually recover but was waiting for the delay threshold to elapse.

  • [major] Ignore NOT_MY_VBUCKET config payloads if CCCP provider is disabled. This allows the client to circumvent any possible bugs in the CCCP response payload and rely entirely on the HTTP config. It also allows 'rewriting' proxies like confsed to function.

2.4.0-beta

  • [major] Better error reporting for SSL failures. This adds new error codes (LCB_SSL_ERROR, LCB_SSL_CANTVERIFY) which are returned on initialization and verification failures respectively.

  • [minor] Communication via legacy memcached servers is possible by using the memcached:// scheme in the connection string.

  • [minor] Environment variables understood by the library are now documented in their own section.

  • [major] Add lcb_get_node() function to retrieve addresses for various nodes in the cluster. This deprecates the lcb_get_host(), lcb_get_port() and lcb_get_server_list() functions as they are constrained to only return information about the administrative API. The new function is configurable to return information about various ports.

  • [major] The dsn field in the lcb_create_st structure has been renamed to connstr.

  • [major] An HTTP request which has followed redirects will cause the lcb_wait() function to never return. This bug was introduced in 2.4.0-DP1 and has now been fixed.

  • [minor] lcb_get_server_list() function now returns updated information from the current cluster configuration. Previously this would only return a node from the list specified during initial creation.

  • [minor] Provide additional error classifiers. Two error classifiers have been added, they are:

    • LCB_ERRTYPE_SRVLOAD which indicates that the server is likely under high load.
    • LCB_ERRTYPE_SRVGEN which indicates that the error is a direct reply from the server. This code can help distinguish between client and server generated return codes.

  • [major] Provide HTTP keepalive and connection pooling for HTTP requests. This allows the client to reuse an HTTP connection for multiple requests rather than creating a new connection and closing it for each operation.

    The functionality may be controlled via the LCB_CNTL_HTTP_POOLSIZE setting which limits how many open connections (per server) to maintain inside the client. Setting this value to 0 will disable pooling and restore old behavior.

  • [major] Properly schedule next invocations for retry queue. A bug was introduced in 2.4.0-dp1 which would cause the next tick callback to be invoked in what is effectively a busy loop. This would be reflected in higher CPU load and less throughput during topology changes.

  • [major] Return error if empty key is passed to an operation. Empty keys will cause the server to drop the connection. The error code returned is the newly added LCB_EMPTY_KEY

  • [minor] Provide setting to disable refreshing the configuration when an HTTP API error is encountered (from one of the HTTP callback functions). This adds the LCB_CNTL_HTTP_REFRESH_CONFIG_ON_ERROR setting.

  • [major] Fix bug where the CCCP provider may prematurely fail, activating the HTTP provider

2.4.0-dp1 (2014-06-18)

Changes affecting older APIs

  • [minor] Make run_event_loop and stop_event_loop private. These functions may no longer be used from within an application to start/stop the event loop. lcb_wait() and lcb_wait3() should be used instead.

  • [major] Deprecate the lcb_set_XXX functions. lcb_set_timeout and some other calls have been deprecated in favor of the lcb_cntl() interface. These functions will still work but will cause the compiler to print a deprecation warning.

  • [minor] lcb_socket_t is typedefed to a DWORD on windows. In previous versions this was an int.

  • [minor] Connecting to a standalone memcached instance is currently no longer supported.

  • [major] lcb_set_error_callback() has been deprecated. Applications should use the new lcb_set_bootstrap_callback() and/or operation callbacks to determine success/failure status.

  • [major] lcb_get_last_error() has been deprecated. Error information is always returned in the operation callback

  • [major] Disable the sending of GETQ packets. The format of this command is cumbersome to deal with and in most uses cases is actually slightly less efficient on the network. Note that this does not change the API of the actual lcb_get() call, but simply changes the format of the packets sent over the wire.

  • [major] The IOPS API has been changed. This is considered volatile interface and may subsequently change in the future as well.

New APIs added in 2.4.0 extending existing functionality

These changes extend existing features with enhanced APIs

  • [major] Additional APIs for lcb_cntl(). These consist of helper functions to make it easier to use simple types or strings rather than pointers, if possible. These functions are lcb_cntl_string(), lcb_cntl_setu32() and lcb_cntl_getu32()

  • [minor] Provide extended version of lcb_wait(). A new function called lcb_wait3() has been added which offers additional options with respect to running the event loop. Specifically it offers to bypass the check for pending operations which lcb_wait() executes. This is both more performant and allows us to wait for operations which are not explicitly scheduled.

  • [major] Provide API to request a configuration refresh. Sometimes it is necessary to force the client to request a new configuration, for example in certain failover conditions. A new API called lcb_config_refresh() has been added, and should be used in conjunction with lcb_wait3().

  • [major] Provide bootstrapping notification callback This provides an explicit lcb_set_bootstrap_callback() to definitively determine whether the client has received its initial configuration (and thus may now start performing operations) or whether it failed (and thus must be reinitialized). This deprecates the common use case of lcb_set_error_callback().

  • [major] New vBucket interface/API. This API is used internally and exposed as volatile inside the public header files. It provides extended features, a more concise API, and is compatible with the upcoming Couchbase 3.0 config format. Note that file-based configuration caches written by this version of the library are incompatible with previous versions, however this version may read caches generated by previous versions. This is because this version generates a stripped-down version of the "terse" configuration style.

  • [major] Extended detailed error codes. These error codes expose more detail about the NETWORK_ERROR and CONNECT_ERROR codes returned by previous versions of the library. The extended codes are not returned by default and must be explicitly enabled in order to retain backwards compatibility with applications which rely on the older error codes.

New Features in 2.4.0

  • [major] Connection Strings (aka "dsn") feature for instance creation. This adds a new version of the lcb_create_st structure which is passed a URI-like string rather than a semicolon-delimited list of hosts. This string is used to provide options and the list of hosts that the library should connect to. For example, couchbase://default/localhost&compression=off

  • [major] SSL transport support for Couchbase 3.0 Enterprise. Couchbase 3.0 enterprise features the ability to encrypt communications between the client and the server using the SSL protocol. SSL protocol support in libcouchbase.

  • [major] Retry queue for failed operations. The retry queue is used as a place to place operations which have failed internally and which should be retried within a certain amount of time. This also provides options on which commands should be retried.

  • [minor] Compression/JSON flag (aka Datatype) support This adds support for a future feature of Couchbase server which will feature transparent compression. This feature also allows the server to signal to the library if a document is JSON or not. The compression feature may be disabled at compile-time, and may also be modified at runtime by setting compression=off in either the connection string or via lcb_cntl_setstring(instance, "compression", "off")

  • [major] Experimental scheduling API. This API replaces most of the older operation APIs with a scheduling API. These APIs are called with one command at a time and insert the resultant packet into a pipeline. The user may "schedule" the commands or "fail" the pipeline if a certain request has failed to be scheduled.

    This API also provides a common ABI header for commands so that they may easily be used via type-punning, or wrapped as a class hierarchy in C++.

    This API is currently considered volatile but will be the basis of the upcoming libcouchbase 3.0 API. The header file is <libcouchbase/api3.h>

  • [major] Raw memcached packets may be sent to the library and have a callback invoked when their responses have been received. This adds an lcb_pktfwd3() API. This requires the new scheduling API.

Bug Fixes in 2.4.0

  • [major] select plugin may endlessly loop in some cases The plugin may loop if there was a long timeout from the future .

  • [major] Do not break TCP connections on topology changes unless ejected from cluster. This ensures that nodes which are still part of the cluster have their TCP connections remain in tact despite being shifted in their server index values. Packets which have been sent to the wrong vBucket are silently ignored and rescheduled to their appropriate destination. This decreases load significantly on the client, network, and cluster during topology changes.

  • [major] Use new-style "Terse" URI format when requesting a configuration over HTTP. This uses the HTTP configuration format over the new /pools/default/bs/default rather than the older /pools/default/bucketsStreaming/default form. The former form is much more efficient on the cluster side. If the new URI form is not supported (i.e. the server responds with an HTTP 404) the older form will be used instead. You may modify this behavior by setting the LCB_CNTL_HTCONFIG_URLTYPE setting via lcb_cntl().

  • [minor] The cmake/configure script now accepts the LDFLAGS, CPPFLAGS, CFLAGS, CXXFLAGS, CC, and CXX settings both within the environment and the commandline, so the forms of CC=clang ./cmake/configure and ./cmake/configure CC=clang are equivalent.

  • [minor] The pillowfight tool will now print latencies between 1-10ms in resolutions of 100us.

Metadata and Packaging Changes in 2.4.0

  • [major] Use Doxygen for API documentation. This replaces the manpages for API documentation with Doxygen. Doxygen is a free and portable documentation system which may be obtained from your distribution or at . To generate the documentation from the source tree, simply run doxygen from the source root directory. To generate internal documentation, run ./docs/gen_internal_apidoc.sh.

  • [major] Add interface attributes to all API calls This properly documents all API calls with a certain API stability level such as committed (for stable APIs), uncomitted for APIs which may, but are not likely to change, and volatile for APIs which are likely to be changed or removed.

  • [major] Public header files have been reorganized This changes the layout of the header files from previous versions. This should not affect applications as applications should only ever include the main <libcouchbase/couchbase.h> file.

    the following files have been removed from the <libcouchbase/*> header directory:

    • types.h - Merged into other header files
    • arguments.h - now a part of couchbase.h
    • callbacks.h - now a part of couchbase.h
    • debug.h - unused and obsolete
    • durability.h - now a part of couchbase.h
    • behavior.h - Merged into deprecated.h
    • sanitycheck.h - Merged into deprecated.h
    • timings.h - Part of couchbase.h
    • compat.h - Part of deprecated.h

    The following files have been added into the <libcouchbase/*> directory. Unless otherwise noted, these files are included by <libcouchbase/couchbase.h>:

    • api3.h - Volatile proposed 3.0 API. Not included by default
    • cxxwrap.h - Contains the implementation for the deprecated C++ wrappers
    • deprecated.h - Contains deprecated APIs
    • iops.h - Contains the IO integration APIs
    • pktfwd.h - Contains the packet forwarding API. Not included by default
    • vbucket.h - Contains the vBucket mapping API. Not included by default

  • OpenSSL is now a base dependency for the library. This may be disabled at configure time via --enable-ssl=no. See ./configure --help.

  • Snappy compression library is bundled and optionally compiled. This is left out by default as the configure script will search for a system installed libsnappy. Snappy provides the compression feature needed for compressing and inflating data between client and server. It may be disabled at compile-time via --enable-snappy=no

  • [minor] libvbucket has been fully integrated into libcouchbase from the forked libvbucket package and, lives fully as part of the library. The public vBucket API may be found in <libcouchbase/vbucket.h>.

  • [minor] As an alternative to the cross-platform lcb_uintNN_t typedefs, a shorter (and more standards compliant) alternative lcb_UNN typedefs are provided, thus instead of lcb_uint32_t you may use lcb_U32. The full listing of cross platform typdefs may be found inside <libcouchbase/sysdefs.h>

2.3.1 (2014-05-08)

  • [major] CCBC-404: Segfault in connmgr_invoke_request Occasionally a segmentation fault would happen when a connection was being released as a result of a connection failure. This was because of invalid list tracking.

  • [major] CCBC-395: Add lcb_cntl() interface for configuration cache Configuration cache options may be set after instantiation using lcb_cntl() with the new LCB_CNTL_CONFIGCACHE operation. The old-style lcb_create_compat creation path is deprecated.

  • [major] CCBC-394: Get-with-replica occasionally crashes on Windows and UV during topology changes. This was due to not allocating a buffer if one did not exist.

  • [major] CCBC-392: ABI compatibility broken between 2.x and 2.3 for lcb_create_compat. This has been fixed by symbol aliasing between versions. Developers are recommended to use the lcb_cntl() API to set the configuration cache, as specified in CCBC-395

  • [major] CCBC-385: Failed assertion on get-with-replica when connection fails. If a connection fails with a CMD_GET_REPLICA command still in the queue an assertion failure will crash the library. This has been fixed by handling the opcode in the failout_single_request function.

  • [major] CCBC-384: Unknown Winsock error codes crash application. This was fixed by providing proper handlers for Winsock codes which were not explicitly converted into their POSIX equivalents.

  • [major] CCBC-376: Fix memory leak in configuration parsing. A leak was introduced in version 2.3.0 by not freeing the JSON pool structure. This has been fixed in 2.3.1

  • [minor] CCBC-370: lcb_get_host and lcb_get_port may return host-port combinations from different servers. If multiple servers are listening on different ports this may result in yielding an invalid endpoint by combining the output from those two functions. This has been fixed in 2.3.1 by returning the host and port from the first host, in lieu of a currently-connected REST endpoint.

  • [minor] CCBC-368: Initial bootstrapping failure may mask LCB_BUCKET_ENOENT calls with LCB_ETIMEDOUT. This has been fixed by not retrying configuration retrieval if an explicit HTTP 404 code is received. Note that when using bootstrap over memcached, a missing bucket may still be manifest as LCB_AUTH_ERROR.

  • [minor] CCBC-367: Ensure lcb_get_host does not return NULL when the associated lcb_t is of LCB_TYPE_CLUSTER. This would cause crashes in some applications which relied on this function to not return NULL.

  • [major] CCBC-389: Fixed Spurious timeouts being delivered in asynchronous use cases. In applications which do not use lcb_wait() the library will potentially time out commands internally triggering an erroneous configuration refresh. While this issue would not end up failing operations it will cause unnecessary network traffic for retrieving configurations. Applications using lcb_wait() are not affected as that function resets the timeout handler.

  • [major] CCBC-332, CCBC-364: Compare configuration revision information for memcached cluster bootstrap. Previously we would refresh the configuration upon receipt of any new configuration update from memcached. This is fixed in 2.3.1 where the configuration will only be applied if it is deemed to be newer than the current configuration. With memcached bootstrap this is only true if the configuration's rev field is higher than the current one.

2.3.0 GA (2014-04-07)

  • [major] CCBC-152: Provide a master-only observe option. This adds a new struct version to the lcb_observe_cmd_t which allows to select only the master node. One can use this to efficiently check if the key exists (without retrieving it). It also allows one to get the CAS of the item without fetching it.

  • [major] CCBC-281: Fix partial scheduling during multi operations. Previously the library would deliver spurious callbacks if multiple operations were scheduled with a single command and one of the operations could not be mapped to a server. This fixes this behavior and ensures that callbacks are only invoked for items if the entire API call succeeded.

  • [major] CCBC-150: Multi-packet commands will no longer deliver spurious callbacks on failure. Previously these commands would be relocated to the same server during a configuration change, resulting in multiple callbacks for the same command. In this case the client would think all the commands had been completed, and when the next response arrived it would incorrectly map it to a different request.

  • [minor] CCBC-327: Fix assumption of vbucket_compare() only returning if a diff exists. This function actually returns a non-NULL pointer always unless it cannot allocate more memory. This bug was introduced with the DP1 release.

  • [minor] CCBC-326: Memcached buckets should use streaming config. This was left unchecked in the DP1 release and has now been fixed.

  • [major] CCBC-351: Enhance performance for configuration parsing. In previous versions receiving multiple configurations at once would cause CPU spikes on slower systems. The configuration parser code has been optimized to alleviate this issue.

  • [minor] CCBC-350: Provide lcb_cntl() API to retrieve the SCM changeset used by the currently loaded binary. This is a more effective way to get the revision as it does not depend on the specific headers the library was compiled with.

  • [major] CCBC-340: Correctly parse "", "0" and "1" for environment variables. In previous versions having the entry set to an empty string or 0 would still be treated by the library as a true value for various environment variables. This has been fixed so that clear "False" values such as the empty string or 0 are treated as such.

2.3.0-dp1 (2014-02-04)

  • [major] CCBC-234: Implementation of [Cluster Configuration Carrier Publication][cccp-wiki]. This is the new and more efficient way to bootstrap from a cluster using the native memcached protocol and is quicker than the previous HTTP bootstrap mechanism, dramatically improving startup times and reducing load on the server. This feature is available in server verions 2.5 and greater. The existing HTTP configuration is still supported and will be employed as a fallback in the event that CCCP is not supported.

    In conjunction with this, a new struct version has been added to the lcb_create_st parameters structure for use with lcb_create. This allows you to get more control over how the client is initialized:

    lcb_t instance; struct lcb_create_st options; lcb_config_transport_t enabled_transports = { LCB_CONFIG_TRANSPORT_CCCP, LCB_CONFIG_TRANSPORT_LIST_END };

    memset(&options, 0, sizeof(options)); options.version = 2; options.v.v2.mchosts = "example.com:11210"; options.v.v2.transports = enabled_transports;

    lcb_error_t rc = lcb_create(&instance, &options); if (rc != LCB_SUCCESS) { fprintf(stderr, "Failed to create instance: %s\n", lcb_strerror(instance, rc)); }

    The above snippet will configure a client to always use the CCCP protocol and never attempt to fall back on HTTP

    The CCCP implementation required a significant rewrite in how sockets were created and re-used. Particularly, a connection pooling feature was implemented.

    Additionally, the cbc command now has an additional -C option which accepts the preferred configuration mechanism to use.

  • [major] CCBC-305: Implement logging hooks.

    This improvements adds various levels of diagnostic logging with the library itself. It may be utilized via the environment (by setting the LCB_LOGLEVEL environment variable to a positive integer -- the higher the number the more verbose the logging).

    Integrators may also use the logging API specified in <libcouchbase/types.h> to proxy the library's logging messages into your own application.

    Current events logged include connection initialization, destruction, connection pool management, configuration changes, and timeouts.

    By default the library is silent.

  • [major] CCBC-316: Allow per-node bootstrap/config timeouts. This change allows more finer grained control over how long to wait per-node to receive updated configuration info. This setting helps adjust the initial and subsequent bootstrap processes to help ensure each node gets a slice of time.

  • [major] CCBC-297: Handle spurious EWOULDBLOCK on UV/Win32 This issue caused odd errors on Windows when large amounts of data would be received on the socket.

2.2.0 (2013-10-05)

  • [major] CCBC-169 Handle 302 redirects in HTTP (views, administrative requests). By default the library will follow up to three redirects. Once the limit reached the request will be terminated with code LCB_TOO_MANY_REDIRECTS. Limit is configurable through LCB_CNTL_MAX_REDIRECTS. If set to -1, it will disable redirect limit.

    int new_value = 5;
lcb_cntl(instance, LCB_CNTL_SET, LCB_CNTL_MAX_REDIRECTS, &new_value);

  • [major] CCBC-243 Replace isasl with cbsasl, the latter has implemented both PLAIN and CRAM-MD5 authentication mechanisms.

    • LCB_CNTL_MEMDNODE_INFO command updated to include effective SASL mechanism:

      cb_cntl_server_t node;
node.version = 1;
node.v.v1.index = 0; /* first node */
lcb_cntl(instance, LCB_CNTL_GET, LCB_CNTL_MEMDNODE_INFO, &node);
if (node.v.v1.sasl_mech) {
    printf("authenticated via SASL '%s'\n",
           node.v.v1.sasl_mech);
}

    • It is also possible to force specific authentication mechanism for the connection handle using LCB_CNTL_FORCE_SASL_MECH command:

      lcb_cntl(instance, LCB_CNTL_SET, LCB_CNTL_FORCE_SASL_MECH, "PLAIN");

  • [major] CCBC-286 libuv plugin: use same CRT for free/malloc

  • [major] CCBC-288 Fail NOT_MY_VBUCKET responses on timeout

  • [major] CCBC-275 Do a full purge when negotiation times out. In this case we must purge the server from all commands and not simply pop individual items.

  • [major] CCBC-275 Reset the server's buffers upon reconnection. This fixes a crash experienced when requesting a new read with the previous buffer still in tact. This was exposed by calling lcb_failout_server on a timeout error while maintaining the same server struct.

  • [major] CCBC-282 Make server buffers reentrant-safe. When purging implicit commands, we invoke callbacks which may in turn cause other LCB entry points to be invoked which can shift the contents and/or positions of the ringbuffers we're reading from.

  • [major] CCBC-204, CCBC-205 Stricter/More inspectable behavior for config cache. This provides a test and an additional lcb_cntl operation to check the status of the configuration cache. Also it switches off config cache with memcached buckets.

    int is_loaded;
lcb_cntl(instance, LCB_CNTL_GET, LCB_CNTL_CONFIG_CACHE_LOADED, &is_loaded);
if (is_loaded) {
    printf("Configuration cache saved us a trip to the config server\n");
} else {
    printf("We had to contact the configuration server for some reason\n");
}

  • [major] CCBC-278 Use common config retry mechanism for bad configcache. This uses the same error handling mechanism as when a bad configuration has been received from the network. New LCB_CONFIG_CACHE_INVALID error code to notify the user of such a situation

  • [major] CCBC-274 Handle getl/unl when purging the server (thanks Robert Groenenberg)

  • [major] Don't failout all commands on a timeout. Only fail those commands which are old enough to have timed out already.

  • [major] CCBC-269 Don't record and use TTP/TTR from observe. Just poll at a fixed interval, as the responses from the server side can be unreliable.

  • [minor] Allow hooks for mapping server codes to errors. This also helps handle sane behavior if a new error code is introduced, or allow user-defined logging when a specific error code is received.

    lcb_errmap_callback default_callback;

lcb_error_t user_map_error(lcb_t instance, lcb_uint16_t in)
{
  if (in == PROTOCOL_BINARY_RESPONSE_ETMPFAIL) {
    fprintf(stderr, "temporary failure on server\n");
  }
  return default_callback(instance, in);
}

...

default_callback = lcb_set_errmap_callback(conn, user_map_error);

  • [minor] Add an example of a connection pool. See example/instancepool directory

  • [minor] CCBC-279 Force lcb_wait return result of wait operation instead of lcb_get_last_error. It returns last_error if and only if the handle is not yet configured

  • [minor] CCBC-284 cbc-pillowfight: compute item size correctly during set If minSize and maxSize are set to the same value it can sometimes crash since it may try to read out of memory bounds from the allocated data buffer.

  • [minor] CCBC-283 Apply key prefix CLI option in cbc-pillowfight

  • [minor] Add --enable-maintainer-mode. Maintainer mode enables --enable-werror --enable-warnings --enable-debug, forces all plugins to be installed and forces all tests, tools, and examples to be built

  • [minor] CCBC-255 Expose LCB_MAX_ERROR to allow user-defined codes

2.1.3 (2013-09-10)

  • [minor] Updated gtest to version 1.7.0. Fixes issue with building test suite with new XCode 5.0 version being released later this month.

  • [major] CCBC-265 Do not try to parse config for LCB_TYPE_CLUSTER handles. It fixes timouts for management operations (like 'cbc bucket-create', 'cbc bucket-flush', 'cbc bucket-delete' and 'cbc admin')

  • [major] CCBC-263 Skip unfinished SASL commands on rebalance. During rebalance, it is possible that the newly added server doesn't have chance to finish SASL auth before the cluster will push config update, in this case packet relocator messing cookies. Also the patch makes sure that SASL command/cookie isn't mixing with other commands

  • [major] Use cluster type connection for cbc-bucket-flush. Although flush command is accessible for bucket type connections, cbc-bucket-flush doesn't use provided bucket name to connect to, therefore it will fail if the bucket name isn't "default".

  • [major] Allow to make connect order deterministic. It allows the user to toggle between deterministic and random connect order for the supplied nodes list. By default it will randomize the list.

  • [major] Do not allow to use Administrator account for LCB_TYPE_BUCKET

  • [major] CCBC-258 Fig segmentation faults during tests load of node.js. Sets inside_handler on socket_connected. Previously we were always using SASL auth, and as such, we wouldn't flush packets from the cmd_log using server_send_packets (which calls apply_want). apply_want shouldn't be called more than once per event loop entry -- so this sets and unsets the inside_handler flag.

  • [major] Added support of libuv 0.8

  • [major] Close config connection before trying next node. It will fix asserts in case of the config node becomes unresponsive, and the threshold controlled by LCB_CNTL_CONFERRTHRESH and lcb_cntl(3)

2.1.2 (2013-08-27)

  • [major] CCBC-253, CCBC-254 Use bucket name in SASL if username omitted. Without this fix, you can may encounter a segmentation faults for buckets, which are not protected by a password.

  • [major] Preserve IO cookie in options_from_info when using v0 plugins with user-provided IO loop instance. This issue was introduced in 2.1.0.

  • [minor] Display the effective IO backend in 'cbc-version'. This is helpful to quickly detect what is the effective IO plugin on a given system.

2.1.1 (2013-08-22)

  • [minor] Use provided credentials for authenticating to the data nodes. With this fix, it is no longer possible to use Administrator credentials with a bucket. If your configuration does so, you must change the credentials you use before applying this update. No documentation guides use of Administrator credentials, so this change is not expected to affect few, if any deployments.

  • [major] CCBC-239 Do not use socket after failout. Fixes segmentation faults during rebalance.

  • [minor] CCBC-245 Distribute debug information with release binaries on Windows

  • [minor] CCBC-248 Do not disable config.h on UNIX-like platforms. It fixes build issue, when application is trying to include plugins from the tarball.

  • [major] CCBC-192 Skip misconfigured nodes in the list. New lcb_cntl(3couchbase) added to control whether the library will skip nodes in initial node list, which listen on configuration port (8091 usually) but doesn't meet required parameters (invalid authentication or missing bucket). By default report this issue and stop trying nodes from the list, like all previous release. Read more at man page lcb_cntl(3couchbase) in section LCB_CNTL_SKIP_CONFIGURATION_ERRORS_ON_CONNECT

  • [major] CCBC-246 Fallback to 'select' IO plugin if default plugin cannot be loaded. On UNIX-like systems, default IO backend is 'libevent', which uses third-party library might be not available at the run-time. Read in lcb_cntl(3couchbase) man page in section LCB_CNTL_IOPS_DEFAULT_TYPES about how to determine effective IO plugin, when your code chose to use LCB_IO_OPS_DEFAULT during connection instantiation. The fallback mode doesn't affect application which specify IO backend explicitly.

2.1.0 (2013-08-17)

  • [major] New backend select. This backend is based on the select(2) system call and its Windows version. It could be considered the most portable solution and is available with the libcouchbase core.

  • [major] CCBC-236 New backend libuv. This backend previously was part of the couchnode project and is now available as a plugin. Because libuv doesn't ship binary packages there is no binary package libcouchbase2-libuv. You can build plugin from the source distribution, or through the libcouchbase-dev or libcouchbase-devel package on UNIX like systems.

  • [major] New backend iocp. This is a Windows specific backend, which uses "I/O Completion Ports". As a part of the change, a new version of plugin API was introduced which is more optimized to this model of asynchronous IO.

  • [major] CCBC-229 Fixed bug when REPLICA_FIRST fails if first try does not return key

  • [major] CCBC-228 Fixed bug when REPLICA_SELECT didn't invoke callbacks for negative error codes

  • [major] CCBC-145 API for durability operations. This new API is based on lcb_observe(3) and allows you to monitor keys more easily. See the man pages lcb_durability_poll(3) and lcb_set_durability_callback(3) for more info.

  • [major] New configuration interface lcb_cntl(3) along with new tunable options of the library and connection instances. In this release the following settings are available. See the man page for more information and examples.:

    • LCB_CNTL_OP_TIMEOUT operation timeout (default 2.5 seconds)

    • LCB_CNTL_CONFIGURATION_TIMEOUT time to fetch cluster configuration. This is similar to a connection timeout (default 5 seconds)

    • LCB_CNTL_VIEW_TIMEOUT timeout for couchbase views (default 75 seconds)

    • LCB_CNTL_HTTP_TIMEOUT timeout for other HTTP operations like RESTful flush, bucket creating etc. (default 75 seconds)

    • LCB_CNTL_RBUFSIZE size of the internal read buffer (default 32768 bytes)

    • LCB_CNTL_WBUFSIZE size of the internal write buffer (default 32768 bytes)

    • LCB_CNTL_HANDLETYPE type of the lcb\_t handler (readonly)

    • LCB_CNTL_VBCONFIG returns pointer to VBUCKET_CONFIG_HANDLE (readonly)

    • LCB_CNTL_IOPS get the implementation of IO (lcb_io_opt_t)

    • LCB_CNTL_VBMAP get vBucket ID for a given key

    • LCB_CNTL_MEMDNODE_INFO get memcached node info

    • LCB_CNTL_CONFIGNODE_INFO get config node info

    • LCB_CNTL_SYNCMODE control synchronous behaviour (default LCB_ASYNCHRONOUS)

    • LCB_CNTL_IP6POLICY specify IPv4/IPv6 policy (default LCB_IPV6_DISABLED)

    • LCB_CNTL_CONFERRTHRESH control configuration error threshold (default 100)

    • LCB_CNTL_DURABILITY_TIMEOUT durability timeout (default 5 seconds)

    • LCB_CNTL_DURABILITY_INTERVAL durability polling interval (default 100 milliseconds)

    • LCB_CNTL_IOPS_DEFAULT_TYPES get the default IO types

    • LCB_CNTL_IOPS_DLOPEN_DEBUG control verbose printing of dynamic loading of IO plugins.

2.0.7 (2013-07-10)

  • [major] CCBC-183 Improve lcb_get_replica(). Now it is possible to choose between three strategies:

    1. LCB_REPLICA_FIRST: Previously accessible and now the default, the caller will get a reply from the first replica to successfully reply within the timeout for the operation or will receive an error.

    2. LCB_REPLICA_ALL: Ask all replicas to send documents/items back.

    3. LCB_REPLICA_SELECT: Select one replica by the index in the configuration starting from zero. This approach can more quickly receive all possible replies for a given topology, but it can also generate false negatives.

    Note that applications should not assume the order of the replicas indicates more recent data is at a lower index number. It is up to the application to determine which version of a document/item it may wish to use in the case of retrieving data from a replica.

2.0.6 (2013-05-07)

  • [major] CCBC-188 Fix segfault when rebalancing When a (!connected) server is reconnected, the tasks in its "pending" buffer will be moved into "output" buffer. If its connection is broken again immediately, relocate_packets() will go to wrong path.

  • [major] CCBC-202 Don't try to switch to backup nodes when timeout is reached

  • [major] CCBC-188 Check if SASL struct is valid before disposing

  • [major] Fix compile error with sun studio "src/event.c", line 172: error: statement not reached (E_STATEMENT_NOT_REACHED)

  • [major] Don't invoke HTTP callbacks after cancellation, because user code might assume a previously-freed resource is still valid

  • [minor] CCBC-179 Added an example to properly use the bucket credentials for authentication instead of administrator credentials

  • [minor] example/yajl/couchview.c: pass cookie to the command Fixes coredump when executing ./examples/yajl/couchview

  • [minor] CCBC-201 Add Host header in http request http://cbugg.hq.couchbase.com/bug/bug-555 points out that Host is a required field in HTTP 1.1

2.0.5 (2013-04-05)

2.0.4 (2013-03-06)

  • [minor] CCBC-185 The bootstrap URI is not parsed correctly

  • [minor] CCBC-175 Work properly on systems where EWOULDBLOCK != EAGAIN

  • [critical] CCBC-180 Segmentation fault when the hostname resolved into several addresses and first of them reject couchbase connections.

  • [major] CCBC-182 The library stops iterating backup nodes list if the next one isn't accessible.

  • [major] CCBC-147 Fixed illegal memory access in win32 plugin

  • [minor] CCBC-178 Build error on solaris/sparc: -Werror=cast-align

2.0.3 (2013-02-06)

  • [minor] bypass SASL LIST MECH

  • [minor] Shrink internal lookup tables (and reduce the size of lcb_t)

  • [minor] Add a new library: libcouchbase_debug.so (see include/libcouchbase/debug.h) which is a new library that contains new debug functionality.

  • [minor] Added manual pages for the library

  • [major] CCBC-153 Reset internal state on lcb_connect(). Allow caller to use lcb_connect() multiple times to implement reconnecting using the same lcb_t instance. Also it sets up the initial-connection timer for users who don't use lcb_wait() and drive IO loop manually.

  • [major] CCBC-171 Invalid read in libevent plugin, when the plugin compiled in 1.x mode

  • [critical] CCBC-155 Observe malfunctions in the case of multiple keys and server failure

  • [major] CCBC-156 The ep-engine renders meaningful body for observe responses only if status code is 0 (PROTOCOL_BINARY_RESPONSE_SUCCESS). We shouldn't interpret response body in other cases, just decode & failout request instead. Also we shouldn't retry observe commands on PROTOCOL_BINARY_RESPONSE_NOT_MY_VBUCKET, because it can cause the client to loop infinitely

  • [major] CCBC-145 KV Durability operation API. Async APIs added to allow the checking of the durability (replication and persistence) status of a key, and to notify the user when a specific criteria has been satisfied.

2.0.2 (2013-01-04)

  • [major] CCBC-150 commands sent to multiple servers fail to detect the respose if mixed with other commands.

  • [minor] CCBC-143 'cbc version' reports that uses 2.0.0, but really installed with 2.0.1. Minor but confusing issue.

  • [major] CCBC-151 Cancellation of the HTTP request might lead to memory leaks or to segfaults (2e3875c2).

  • [minor] Document LCB_SERVER_BUG and LCB_PLUGIN_VERSION_MISMATCH. Enhance the the lcb_strerror() test to detect undocumented error codes.

  • [critical] CCBC-153 Under high load the library could generate LCB_ETIMEDOUT errors without reason owing to internal limitations.

2.0.1 (2012-12-11)

50 files changed, 1009 insertions(+), 274 deletions(-)

  • libev-plugin: delay all timers while the loop isn't active. It will fix LCB_ETIMEOUT in the following scenario:

    • connect the instance
    • sleep for time greater than default timeout (e.g. 3 seconds)
    • schedule and execute a command (it will be timed out immediately)

  • libev-plugin: reset IO event on delete. We need to reset it, because it might be re-used later

  • CCBC-136: do not abort when purging SASL commands

  • Make library C89 friendly again

  • CCBC-132, CCBC-133: Ensure HTTP works even when the network may be unreliable. This changeset encompasses several issues which had been found with HTTP requests during network errors and configuration changes. Specifically some duplicate code paths were removed, and the process for delivering an HTTP response back to the user is more streamlined.

  • CCBC-130: Fix a memory leak on the use of http headers

  • CCBC-131: Compensate for cluster nodes lacking couchApiBase

  • Fix possible SEGFAULT. Not-periodic timers are destroyed after calling user's callback, after that library performed read from freed pointer.

  • SystemTap and DTrace integration

2.0.0 (2012-11-27)

12 files changed, 50 insertions(+), 12 deletions(-)

  • Install unlock callback in synchronous mode

  • Add the CAS to the delete callback

  • Minor update of the packaging layout:

    • libcouchbase-all package comes without version
    • extract debug symbols from libcouchbase-{bin,core} to libcouchbase-dbg package

2.0.0beta3 (2012-11-21)

64 files changed, 3641 insertions(+), 735 deletions(-)

  • CCBC-104 Fix illegal memory access. Reconnect config listener if the config connection was gone without proper shutdown.

  • Check for EWOULDBLOCK/EINTR on failed send

  • Allow to use gethrtime() from C++

  • Fix using freed memory (was introduced in 4397181)

  • Use dynamic versioning for plugins

  • Remove libtool version from the plugins

  • Allow to use 'cbc-hash' with files

  • CCBC-120 Purge stale OBSERVE packets

  • CCBC-120 Reformat and refactor lcb_server_purge_implicit_responses:

    • move packet allocation out of GET handler
    • dropping NOOP command shouldn't return error code

  • CCBC-122 Try to switch another server from backup list on timeout

  • CCBC-119: Allow the user to specify a different hash key. All of the data operations contains a hashkey and nhashkey field. This allows you to "group" items together in your cluster. A typical use case for this is if you're storing lets say data for a single user in multiple objects. If you want to ensure that either all or none of the objects are available if a server goes down, it could be a good idea to locate them on the same server. Do bear in mind that if you do try to decide where objects is located, you may end up with an uneven distribution of the number of items on each node. This will again result in some nodes being more busy than others etc. This is why some clients doesn't allow you to do this, so bear in mind that by doing so you might not be able to get your objects from other clients.

  • Create man pages for cbc and cbcrc

  • CCBC-118 lcb_error_t member in the http callbacks shouldn't reflect the HTTP response code. So the error code will be always LCB_SUCCESS if the library managed to receive the data successfully.

  • Timer in libev uses double for interval. Ref:

  • CCBC-115 Return zero from do_read_data() if operations_per_call reached. The `operations_per_call' limit was introduced to prevent from freezing event loop. But in the function variable rv could store two different results and in case of reaching this limit it is returning number of the processed records, which is wrong. The function should return either zero (success) or non-zero (failure).

  • Do not allow admin operations without authentication

  • Fix cbc-bucket-create. `sasl-password' is misspelled, and it fails to parse the command line option.

  • CCBC-114 Lookup the plugin symbol also in the current executable image.

  • CCBC-113 Remove unauthorized asserion (d344037). The lcb_server_send_packets() function later check if the server object connected and establish connection if not (with raising possible errors)

  • Try all known plugins for LCB_IO_OPS_DEFAULT in run time

  • Don't use the time_t for win32. When compiling from php it turns out that it gets another size of the time_t type, causing the struct offsets to differ.

  • Add lcb_verify_compiler_setup(). This function allows the "user" of the library to verify that the compiler use a compatible struct packing scheme.

  • CCBC-87: Add documentation about the error codes

2.0.0beta2 (2012-10-12)

81 files changed, 2822 insertions(+), 1353 deletions(-)

  • Search ev.h also in ${includedir}/libev

  • Fix SEGFAULT if IO struct is allocated not by the lcb_create()

  • Allow libcouchbase to connect to an instance without specifying bucket. It is useful when the bucket not needed, e.g. when performing administration tasks.

  • Fix memory leak after an unsuccessful connection

  • Fix invalid memory access in cbc tool. Affected command is cbc-bucket-create

  • lcb_create: replace assert() with error code

  • CCBC-105 breakout event loop in default error_callback. This provides better default behaviour for users who haven't defined global error callback.

  • Allow users to build the library without dependencies. For example, without plugins at all. This may be useful if the plugin is implemented by or built into the host application.

  • Allow users to install both libraries (2.x and 1.x) on the same system.

  • Make the content type optional for lcb_make_http_request()

  • Fix password memory leak in http.c (7e71493)

  • Add support for raw http requests. libcouchase already contains all the bits to execute a raw http request, except for the possibility to specify a host:port, username and password.

  • Cleanup HTTP callbacks. Use the same callbacks both for Management and View commands, and rename them to lcb_http_complete_callback and lcb_http_data_callback.

  • Allow users to use environment variables to pick the event plugin

  • Add a new interface version for creating IO objects via plugins

  • Implement a new libev plugin. It is compatible with both libev3 and libev4.

  • CCBC-103: Fix linked event/timer lists for win32

  • Allow to disable CXX targets

  • lcb_connect() should honor the syncmode setting. Automatically call lcb_wait() when in synchronous mode

2.0.0beta (2012-09-13)

123 files changed, 13753 insertions(+), 8264 deletions(-)

  • Refactor the API. This is a full redesign of the current libcouchbase API that'll allow us to extend parts of the API without breaking binary compatibility. Also it renames all functions to have lcb_ prefix instead of libcouchbase_ and LCB/LIBCOUCHBASE in macros.

  • Added --enable-fat-binary. Helps to solve issues when linking with fat binaries on MacOS.

  • Implement getter for number of nodes in the cluster: lcb_get_num_nodes()

  • Implement RESTful flush in the cbc toolset

  • Bundle Windows packages as zip archives

  • CCBC-98 Differentiate between TMPFAILs. This allows a developer to know if the temporary condition where the request cannot be handled is due to a constraint on the client or the server.

  • Don't try to put the current node last in the backup list. This may cause "duplicates" in the list if the REST server returns another name for the server than you used. Ex: you specify "localhost" and the REST response contains 127.0.0.1

  • Fix locking keys in multi-get mode

  • Fix bug where HTTP method is not set

  • CCBC-96 Correct buffer length for POST/PUT headers

  • Add lcb_get_server_list

  • Merge lcb_get_locked into lcb_get function

  • Fix Windows build

  • Include sys/uio.h. Needed by OpenBSD

  • Fix mingw build (c394a1c)

  • CCBC-80: Default to IPv4 only

  • Sync memcached/protocol_binary.h. Pull extra protocol_binary_datatypes declarations.

  • Deliver HTTP headers via callbacks

  • Unify HTTP interface. This means massive rename of the symbols

  • CCBC-92 release ringbuffer in lcb_purge_single_server

  • CCBC-91 Fix switching to backup node in case of server outage

  • CCBC-91 Reset timer for commands with NOT_MY_VBUCKET response

  • Fix alignment for sparc platforms

  • Fix win32 build (Add strings.h)

  • Fix build with libyajl available

  • Bundle libvbucket

  • Fix a problem with allocating too few slots in the backup_nodes. Fixes illegal memory access.

  • CCBC-90 Fix initialization of backup nodes array. The code switching nodes relies on NULL terminator rather than nbackup_nodes variable. Fixes illegal memory access.

  • CCBC-89: Release the memory allocated by the http parser

1.0.6 (2012-08-30)

5 files changed, 18 insertions(+), 5 deletions(-)

  • CCBC-92 release ringbuffer in libcouchbase_purge_single_server

1.0.5 (2012-08-15)

6 files changed, 23 insertions(+), 15 deletions(-)

  • CCBC-91 Fix switching to backup node in case of server outage

  • CCBC-91 Reset timer for commands with NOT_MY_VBUCKET response

1.1.0dp9 (2012-07-27)

5 files changed, 18 insertions(+), 11 deletions(-)

  • Render auth credentials for View requests. libcouchbase_make_http_request() won't accept credentials anymore. It will pick them bucket configuration.

1.1.0dp8 (2012-07-27)

36 files changed, 2093 insertions(+), 704 deletions(-)

  • Allow the client to specify the verbosity level on the servers using lcb_set_verbosity() function.

  • Bind timeouts to server sockets instead of commands. This means that from this point timeout interval will be started from the latest IO activity on the socket. This is a behavior change from the 1.0 series.

  • Allow the user to get the number of replicas using libcouchbase_get_num_replicas()

  • Allow a user to breakout from the event loop in callbacks using libcouchbase_breakout()

  • Make libcouchbase_wait() re-entrable

  • Let users detect if the event loop running already using libcouchbase_is_waiting() function.

  • CCBC-77 Use separate error code for ENOMEM on the client

  • CCBC-82 Implement read replica

  • CCBC-85 Implement general purpose timers. It is possible for users to define their own timers using libcouchbase_timer_create() function. (See headers for more info)

  • Implement multiple timers for windows

  • CCBC-15 Add OBSERVE command

  • Allow users to specify content type for HTTP request

  • Fix to handle the case when View base doesn't have URI schema

  • Separate HTTP callbacks for couch and management requests

  • Claim that server has data in buffers if there are HTTP requests pending. Without this patch the event loop can be stopped prematurely.

  • Add new cbc commands and options:

    • cbc-view (remove couchview example)
    • cbc-verbosity
    • cbc-admin
    • cbc-bucket-delete
    • cbc-bucket-create
    • Add -p and -r options to cbc-cp to control persistence (uses OBSERVE internally)

1.1.0dp7 (2012-06-19)

18 files changed, 266 insertions(+), 115 deletions(-)

  • Add support for notification callbacks for configuration changes. Now it is possible to install a hook using function libcouchbase_set_configuration_callback(), and be notified about all configuration changes.

  • Implement function to execution management requests. Using libcouchbase_make_management_request() function you can configure the cluster, add/remove buckets, rebalance etc. It behaves like libcouchbase_make_couch_request() but works with another endpoint.

  • Extract HTTP client. Backward incompatible change in Couchbase View subsystem

1.1.0dp6 (2012-06-13)

20 files changed, 201 insertions(+), 127 deletions(-)

  • CCBC-70 Close dynamic libraries. Fixes small memory leak

  • CCBC-72 Fix compilation on macosx with gtest from homebrew

  • CCBC-71 Implement 'help' command for cbc tool

  • Undefine NDEBUG to avoid asserts to be optimized out

  • Fix win32 builds:

    • Add suffix to cbc command implementations
    • Fix guards for socket errno macros
    • Define size_t types to fix MSVC 9 build
    • MSVC 9 isn't C99, but has stddef.h, so just include it

  • CCBC-63 Include types definitions for POSIX systems. Fixes C++ builds on some systems.

1.1.0dp5 (2012-06-06)

7 files changed, 65 insertions(+), 9 deletions(-)

  • The library doesn't depend on pthreads (eliminates package lint warnings)

  • Implement 'cbc-hash' to match server/vbucket for given key

1.1.0dp4 (2012-06-05)

8 files changed, 54 insertions(+), 7 deletions(-)

  • cbc: strtoull doesn't exist on win32, therefore use C++ equivalent.

  • integration with Travis-CI

1.1.0dp3 (2012-06-03)

54 files changed, 1874 insertions(+), 824 deletions(-)

  • CCBC-68 Implement UNLOCK_KEY (UNL) command

  • CCBC-68 Implement GET_LOCKED (GETL) command

  • hashset.c: iterate over whole set on rehashing. Fixes memory leaks related to hash collisions (905ef95)

  • Destroy view requests items when server get destroyed

  • Do not call View callbacks for cancelled requests

  • Fix ringbuffer_memcpy() (36afdb2)

  • CCBC-62 A hang could occur in libcouchbase_wait() after the timeout period. Check for breakout condition after purging servers

  • CCBC-65 A small memory leak can occur with frequent calls to libcouchbase_create() and libcouchbase_destroy()

  • CCBC-64. Timeouts can occur during topology changes, rather than be correctly retried. Send the retry-packet to new server

  • vbucket_found_incorrect_master() returns server index

  • Fix ringbuffer_is_continous()

  • Pick up cookies from pending buffer unless node connected

  • RCBC-33 A fix for a buffer overflow with the supplied password as has been integrated. While it is a buffer overflow issue, this is not considered to be a possible security issue because the password to the bucket is not commonly supplied by an untrusted source

1.0.4 (2012-06-01)

15 files changed, 330 insertions(+), 76 deletions(-)

  • CCBC-65 A small memory leak can occur with frequent calls to libcouchbase_create() and libcouchbase_destroy()

  • CCBC-62 A hang could occur in libcouchbase_wait() after the timeout period. Check for breakout condition after purging servers

  • CCBC-64. Timeouts can occur during topology changes, rather than be correctly retried. Send the retry-packet to new server

  • [backport] vbucket_found_incorrect_master() returns server index. (orig: c32fdae)

1.0.3 (2012-05-02)

6 files changed, 44 insertions(+), 7 deletions(-)

  • [backport] Fix ringbuffer_is_continous() (orig: 9cfda9d)

  • [backport] Pick up cookies from pending buffer unless node connected (orig: 463958d)

  • RCBC-33 A fix for a buffer overflow with the supplied password as has been integrated. While it is a buffer overflow issue, this is not considered to be a possible security issue because the password to the bucket is not commonly supplied by an untrusted source

1.1.0dp2 (2012-04-10)

10 files changed, 54 insertions(+), 20 deletions(-)

  • CCBC-59 Don't wait for empty buffers. If called with no operations queued, libcouchbase_wait() will block forever. This means that a single threaded application that calls libcouchbase_wait() at different times to make sure operations are sent to the server runs the risk of stalling indefinitely. This is a very likely scenario.

  • Don't define size_t and ssize_t for VS2008

  • Fix segfault while authorizing on protected buckets (211bb04)

1.1.0dp (2012-04-05)

59 files changed, 4374 insertions(+), 1205 deletions(-)

  • This release adds new functionality to directly access Couchbase Server views using the libcouchbase_make_couch_request() function. See the associated documentation and header files for more details

  • Check for newer libvbucket

  • MB-4834: Request the tap bytes in a known byte order (adf2b30)

1.0.2 (2012-03-06)

83 files changed, 4095 insertions(+), 654 deletions(-)

  • Implement VERSION command from binary protocol

  • Allow use of libcouchbase to pure memcached clusters by using libcouchbase_create_compat() function

  • Always sign deb packages and allow to pass PGP key

  • Bundle the protocol definitions for memcached (memcached/protocol_binary.h and memcached/vbucket.h) to make it easier to build

  • Bundle sasl client implementation

  • Fix windows build for MS Visual Studio 9

    • define E* if missing
    • stdint header

  • Add support for multiple hosts for the bootstrap URL. A list of hosts:port separated by ';' to the administration port of the couchbase cluster. (ex: "host1;host2:9000;host3" would try to connect to host1 on port 8091, if that fails it'll connect to host2 on port 9000 etc)

  • Raise error if <stdint.h> missing

  • Add JSON support for cbc-cp command

  • Add option to set timeout for cbc

  • Added support for '-' to cp

  • Added cbc-verify: verify content in cache with files

  • Now cbc supports better usage messages

1.0.1 (2012-02-13)

65 files changed, 3275 insertions(+), 1329 deletions(-)

  • CCBC-38 Use alternate nodes when current is dead. A fix to allow the client library to failover automatically to other nodes when the initial bootstrap node becomes unavailable has been added. All users are recommended to upgrade for this fix.

  • Fix connect timeouts. Timeouts are per-operation and only set if there is any I/O. The special exception to this is initial connections, which do not necessarily have a data stream or write buffer associated wiht them yet.

  • Update to new MT-safe libvbucket API

  • Add option for embedding libevent IO plugin

  • Fix multi-{get,touch} requests handling when nkeys > 1

  • Allow to build without tools which require C++ compiler

  • Destroy event base if we created it

  • CCBC-51 Check server index before using

  • Handle PROTOCOL_BINARY_RESPONSE_NOT_MY_VBUCKET and retry it until success or another error, which can be handled by caller

  • Do not attempt SASL when SASL already in progress

  • Finer grained error reporting for basic REST errors:

    • return LIBCOUCHBASE_AUTH_ERROR on HTTP 401
    • return LIBCOUCHBASE_BUCKET_ENOENT on HTTP 404
    • event loop is stopped (via maybe_breakout) on REST error

  • Fixed segfaults and memory access errors on libevent1.4

  • Allow for notification on initial vbucket config. This makes libcouchbase stop_event_loop and libcouchbase maybe breakout work properly in cooperative asynchronous event loops. the wait flag is set by libcouchbase_wait() and unset by maybe_breakout. Additionally, breakout_vbucket_state_listener will call maybe_breakout as well, instead of having synchronous behavior assumed by libcouchbase_wait()

  • Fix sasl_list_mech_response_handler(). sasl_client_start() expects null-terminated string

  • Refactor: use libcouchbase_xxxx for the datatypes

  • Do not notify user about the same error twice. Use command callback when it's possible. (e.g. where the libcouchbase_server_t is accessible and we can libcouchbase_failout_server())

  • Install configuration.h for win32

  • CCBC-20 Implement operation timeouts. Timeouts applies for all operations, and the timer starts running from the moment you call the libcouchbase operation you want. The timer includes times for connect/send/ receive, and all of the time our application spend before letting the event loop call callbacks into libcouchbase.

  • Fix double free() error when reading key from packet in handler.c (b5d485a)

1.0.0 (2012-01-22)

170 files changed, 6048 insertions(+), 7553 deletions(-)

  • Allow the user to specify sync mode on an instance

  • Empty string as bucket name should be treated as NULL

  • Bail out if you can't find memcached/vbucket.h and libvbucket/vbucket.h

  • New command cbc. This command intended as the analog of mem* tools from libmemcached distribution. Supported commands:

    • cbc-cat
    • cbc-cp
    • cbc-create
    • cbc-flush
    • cbc-rm
    • cbc-stats
    • cbc-send
    • cbc-receive

  • CCBC-37 allow config for cbc tool to be read from .cbcrc

  • Convert flags to network byte order

  • Remove <memcached/vbucket.h> dependency

  • Use the error handler instead of printing to stderr

  • Disable Views code

  • Don't accept NULL as a valid "callback"

  • Add make targets to build RPM and DEB packages

  • Allow download memcached headers from remote host

  • Added docbook-based manual pages

  • Gracefully update vbucket configuration. This means that the connection listener, could reconfigure data sockets on the fly

  • Allow libcouchbase build with libevent 1.x (verified for 1.4.14)

  • Aggregate flush responses

  • Add stats command

0.3.0 (2011-11-02)

102 files changed, 6188 insertions(+), 1531 deletions(-)

  • Add flush command from binary protocol

  • Remove packet filter

  • Use ringbuffers instead buffer_t

  • Win32 build fixes

  • Allow to specify IO framework but using IO plugins

  • CCBC-11 The interface to access views

  • Initial man pages

  • Extend the test suite

0.2.0 (2011-09-01)

85 files changed, 12144 insertions(+)

  • Simple bootstapping which builds HTTP packet and listens /pools/default/buckets/BUCKETNAME directly. Allowed usage of defaults (bucket name, password)

  • Support basic set of binary protocol commands:

    • get (get and touch)
    • set
    • increment/decrement
    • remove
    • touch

  • MB-3294 Added _by_key functions

  • CCBC-5 Fixed abort in do_read_data (c=0x7b09bf0) at src/event.c:105

  • Added timings API. It might be possible to turn on timing collection using libcouchbase_enable_timings()/libcouchbase_disable_timings(), and receive the data in timings callback.

  • Basic TAP protocol implementation

  • Initial win32 support