[ 
https://issues.apache.org/jira/browse/HDDS-7915?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Ethan Rose resolved HDDS-7915.
------------------------------
    Resolution: Fixed

> Force close QUASI_CLOSED replicas only when the container is CLOSED in Legacy 
> RM
> --------------------------------------------------------------------------------
>
>                 Key: HDDS-7915
>                 URL: https://issues.apache.org/jira/browse/HDDS-7915
>             Project: Apache Ozone
>          Issue Type: Sub-task
>          Components: SCM
>            Reporter: Siddhant Sangwan
>            Assignee: Siddhant Sangwan
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.4.0
>
>
> LegacyReplicationManager#closeReplicasIfPossible tries to close replicas. 
> {code}
>   private int closeReplicasIfPossible(ContainerInfo container,
>                                       List<ContainerReplica> replicas) {
>     // This method should not be used on open containers.
>     if (container.getState() == LifeCycleState.OPEN) {
>       return 0;
>     }
>     int numCloseCmdsSent = 0;
>     Iterator<ContainerReplica> iterator = replicas.iterator();
>     while (iterator.hasNext()) {
>       final ContainerReplica replica = iterator.next();
>       final State state = replica.getState();
>       if (state == State.OPEN || state == State.CLOSING) {
>         sendCloseCommand(container, replica.getDatanodeDetails(), false);
>         numCloseCmdsSent++;
>         iterator.remove();
>       } else if (state == State.QUASI_CLOSED) {
>         // Send force close command if the BCSID matches
>         if (container.getSequenceId() == replica.getSequenceId()) {
>           sendCloseCommand(container, replica.getDatanodeDetails(), true);
>           numCloseCmdsSent++;
>           iterator.remove();
>         }
>       }
>     }
>     return numCloseCmdsSent;
>   }
> {code}
> In one case, this method is called from 
> LegacyReplicationManager#handleUnderReplicatedUnhealthy, where the replica's 
> state matches the container state. If the state is QUASI_CLOSED, we will end 
> up closing the replica and eventually the container, even though the replica 
> doesn't have all the data. This should not happen and is what this jira is 
> meant to fix. Force closing a QUASI_CLOSED replica of a QUASI_CLOSED 
> container is handled separately earlier in the processContainer method.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to