Releases: cloudfoundry/nfs-volume-release
Releases · cloudfoundry/nfs-volume-release
nfs-volume v1.7.0
Release Notes
- Adds a db.skip_hostname_validation property to the nfsbrokerpush job to allow TLS connection to databases without validating that the hostname matches the server TLS certificate.
- Adds a db_skip_hostname_validation property to the nfsbroker-bbr job to allow backups and restores with TLS connections and no hostname validation.
Known Issues
- This release contains an issue in nfsbroker causing new nfs service bindings to use the nfs-legacy mounter, potentially causing slower performance and failures when the file server only supports nfs v4. This issue is fixed in v1.7.6 and later releases.
- Due to a bug in monit scripting, the nfsv3driver job may continue to run when it should be stopped. This is corrected in v1.7.1.
- As of version 1.5.2, the
nfsbrokerpush
errand requires thecf-cli-6-linux
job from cf-cli-release v1.6.0 to be colocated on the errand VM. - This version requires the
mapfs
job from mapfs-release v1.1.1 to be colocated on the Diego cell. Using older mapfs releases will cause the nfsv3driver job to fail with a missing bosh link. - This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
nfs-volume v1.6.1
Release Notes
- Improves drain and unmount logic to allow nfsv3driver to safely clean up mounts that are busy or disconnected.
- Improves log format to default logging to human readable timestamp format.
- Updates nfsbroker application manifest bosh template to fix property deprecation warnings and remove unnecessary properties.
- Adds OBSAPI service metadata properties to nfsbroker catalog.
Known Issues
- This release contains an issue in nfsbroker causing new nfs service bindings to use the nfs-legacy mounter, potentially causing slower performance and failures when the file server only supports nfs v4. This issue is fixed in v1.7.6 and later releases.
- Due to a bug in monit scripting, the nfsv3driver job may continue to run when it should be stopped. This has been corrected in v1.7.1.
- As of version 1.5.2, the
nfsbrokerpush
errand requires thecf-cli-6-linux
job from cf-cli-release v1.6.0 to be colocated on the errand VM. - This version requires the
mapfs
job from mapfs-release v1.1.1 to be colocated on the Diego cell. Using older mapfs releases will cause the nfsv3driver job to fail with a missing bosh link. - This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
nfs-volume v1.6.0
Release Notes
- README updated to remove redundant usage notes also found in CF docs.
- Removed temporary connection close code we added to work around a voldriver bug in diego release (long since fixed).
- Fixed drain logic in nfsv3driver that could cause application data loss in nfs-experimental mounts
- Use unique volume mountpoints for each application container to resolve reference counting issues that could cause volume mounts to fail when scaling apps up and down under load.
- upgrade to Go 1.11 for package compilation
- remove redundant ops file that are now available in cf-deployment
- rename
nfs-experimental
service tonfs
in the catalog - rename
nfs
service tonfs-legacy
and deprecate - add experimental support for using CredHub as an nfsbroker backing store instead of a SQL database
- remove vestigial consul DNS job from nfstestserver ops file
Known Issues
- Due to a bug in monit scripting, the nfsv3driver job may continue to run when it should be stopped. This has been corrected in v1.7.1.
- As of version 1.5.2, the
nfsbrokerpush
errand requires thecf-cli-6-linux
job from cf-cli-release v1.6.0 to be colocated on the errand VM. - As of version 1.3.1, the
nfsv3driver
job requires themapfs
job from mapfs-release v1.0.1 to be colocated on the Diego cell. - This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
nfs-volume v1.2.3
- Fixes unsafe logic in unmount and drain code that may lead to deletion of files on remote nfs shares
nfs-volume v1.5.3
Release Notes
- adds logic to the
nfsbrokerpush
BOSH errand to detect whether or not thecflinuxfs3
stack is available in Cloud Foundry and explicitly pushes with-s cflinuxfs3
if it is. - improves error handling for
nfs-experimental
service instances so that mount errors are propagated to cf logs when an app fails to start with a mount failure.
Known Issues
- As of version 1.5.2, the
nfsbrokerpush
errand requires thecf-cli-6-linux
job from cf-cli-release v1.6.0 to be colocated on the errand VM. - As of version 1.3.1, the
nfsv3driver
job requires themapfs
job from mapfs-release v1.0.1 to be colocated on the Diego cell. - This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
nfs-volume v1.5.2
Release Notes
- LDAP uid/gid resolution now supports LDAP servers that have blank
gidNumber
entries by settinggid
equal touid
in this case. - experimental support for client credentials authentication in our BOSH errands. (This only partially works because clients authenticated with OAUTH client credentials cannot yet create cf orgs or spaces.)
- remove cf cli package from nfs-volume-release in favor of colocating a job from cf-cli-release.
Known Issues
- As of version 1.5.2, the
nfsbrokerpush
errand requires thecf-cli-6-linux
job from cf-cli-release v1.6.0 to be colocated on the errand VM. - As of version 1.3.1, the
nfsv3driver
job requires themapfs
job from mapfs-release v1.0.1 to be colocated on the Diego cell. - This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
nfs-volume v1.5.1--DON'T USE
this prerelease includes untested commits. Please don't use it.
nfs-volume v1.5.0
Release Notes
- Performance improvements for readonly mounts using the nfs-experimental service
- Performance improvements for nfs-experimental services that don’t set uid and gid parameters
- Improved data safety in unmount and mount point cleanup code
- cf-deployment 3.0 support in ops files and ci test scripts
Known Issues
- This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
- As of version 1.3.1, the
nfsv3driver
job requires themapfs
job from mapfs-release v1.0.1 to be colocated on the Diego cell.
nfs-volume v1.2.2
- read-only mounts using the nfs-experimental service use default attribute caching instead of disabling it
- nfs-experimental mounts that don't specify uid and gid in the service binding skip mapfs mounting and expose nfs kernel mounts directly without any uid mapping.
nfs-volume v1.4.1
Release Notes
- Added support for cf-deployment 2.0 with PXC databases
- Added namespacing for service broker store data when stored to CredHub (to support future migration/export use cases)
- Added documentation for PXC support, CredHub support, and errands for broker push and broker registration
- Added migration errand to move existing service broker records from SQL to CredHub
- Fixed default service ID nfs-experimental service to match older nfs-volume-release versions, so that we don’t encounter service broker upgrade problems migrating from versions <= 1.3.1
Known Issues
- This release is incompatible with garden-runc-release < 1.3.0. Using it with an older garden release will cause read-only NFS mounts to fail.
- As of version 1.3.1, the
nfsv3driver
job requires themapfs
job from mapfs-release v1.0.1 to be colocated on the Diego cell.