NPE at ClusterBackupAgent.reset

[artem-v]

Hi

Aeron 1.44.1

1 observations from 2024-08-22 15:05:16.554+0300 to 2024-08-22 15:05:16.554+0300 for:
io.aeron.cluster.client.ClusterException: WARN - failed to stop log replay
        at io.aeron.cluster.ClusterBackupAgent.reset(ClusterBackupAgent.java:368)
        at io.aeron.cluster.ClusterBackupAgent.resetBackup(ClusterBackupAgent.java:647)
        at io.aeron.cluster.ClusterBackupAgent.doWork(ClusterBackupAgent.java:287)
       ...
Caused by: java.lang.NullPointerException: Cannot invoke "io.aeron.archive.client.AeronArchive.stopReplay(long)" because "this.clusterArchive" is null
        at io.aeron.cluster.ClusterBackupAgent.reset(ClusterBackupAgent.java:364)
        ... 9 more

Pretty regularly receiving such error in the counted error log file. Looks like there're conditions on which clusterArchive is null during .reset() call in the ClusterBackupAgent.

Unfortunately cannot post clear test that reproduce this.

[vyazelenko]

This could happen if ClusterBackup goes through the following state transition BACKING_UP -> BACKUP_QUERY -> RESET_BACKUP, i.e. the ClusterBackup.SourceType validation failure (e.g. connecting to the leader node while ClusterBackup.SourceType#FOLLOWER is used).
I've pushed a fix c37546e.