Generate Unique EcChronos ID #678 #713
Conversation
SajidRiaz138
requested review from
tommystendahl,
VictorCavichioli and
jwaeab
- Generate eccrhonos_Id that should be unique (uuid + host name) - on start up instance, first check nodes_metadata table for ecchronosId, if it already exists, use that one, otherwise generate now one. - So due to persistence of eccronosId, it would be same even instance get reinstalled or restart. - when instance try to acquire nodes then add this newly ecchronosId into nodes_metadata table with datacenter and nodeId. - The new table nodes_metadata schenma, datacenter is partition key and nodeId is clustering colum and ecchrons_id coulum would hold the Id.
| @@ -64,7 +64,6 @@ public class BeanConfigurator | |||
| private final AtomicReference<Security.CqlSecurity> cqlSecurity = new AtomicReference<>(); | |||
| private final AtomicReference<Security.JmxSecurity> jmxSecurity = new AtomicReference<>(); | |||
| private final ConfigRefresher configRefresher; | |||
| private final String ecChronosID; | |||
There was a problem hiding this comment.
Idk about removing it, because we need to generate this ans pass for EccNodeMetadata if the instance is new, I think here we need to have some validation using the EccNodeMetadata, generate the unique id, and them passing for EccNodesSync
| import static com.datastax.oss.driver.api.querybuilder.QueryBuilder.bindMarker; | ||
| import static com.datastax.oss.driver.api.querybuilder.QueryBuilder.selectFrom; | ||
|
|
||
| public final class EccNodeMetadata |
| public EccNodeMetadata(final List<Node> nodes, final CqlSession session) | ||
| { | ||
| mySession = session; | ||
| myNode = nodes.stream().findAny().get(); |
There was a problem hiding this comment.
What happens if the topology changes and you get a node that was not on the first configuration for a instance that was working?
let's say that in the configuration we are using datacenterAware, so we have a datacenter list of [d1, dc2] but now the topology changed and we added dc3 on the list, but it was not on the previously topology for this instance, so if this line picks up a node from dc3 it will generate a new ecchronos id for a instance that already has an ID
here is first idea to generate and persist unique id for ecchronos id that is partition key for other nodes_sync table.