[
https://issues.apache.org/jira/browse/CASSANDRA-20510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17940730#comment-17940730
]
Stefan Miklosovic commented on CASSANDRA-20510:
-----------------------------------------------
It boils down to this in StorageService
{code}
public UUID getLocalHostUUID()
{
// Metadata collector requires using local host id, and flush of
IndexInfo may race with
// creation and initialization of cluster metadata service. Metadata
collector does accept
// null localhost ID values, it's just that TokenMetadata was created
earlier.
ClusterMetadata metadata = ClusterMetadata.currentNullable();
if (metadata == null ||
metadata.directory.peerId(getBroadcastAddressAndPort()) == null)
return null;
return metadata.directory.peerId(getBroadcastAddressAndPort()).toUUID();
}
{code}
in SimpleSSTableMultiWriter#create method there is
{code}
MetadataCollector metadataCollector = new
MetadataCollector(metadata.get().comparator)
.commitLogIntervals(commitLogPositions != null ? commitLogPositions :
IntervalSet.empty())
.sstableLevel(sstableLevel);
{code}
which uses this constructor:
{code}
public MetadataCollector(ClusteringComparator comparator)
{
this(comparator, StorageService.instance.getLocalHostUUID());
}
{code}
which returns null and then in CommitLogReplayer#persistedIntervals it gets it
as null, hm ...
So, CommitLogReplayer#persistedIntervals is executed sooner than metadata are
initialized as it returns null only in that case.
> Investigate uknown sstables warning in CommitLogReplayer
> --------------------------------------------------------
>
> Key: CASSANDRA-20510
> URL: https://issues.apache.org/jira/browse/CASSANDRA-20510
> Project: Apache Cassandra
> Issue Type: Bug
> Reporter: Stefan Miklosovic
> Priority: Normal
>
> This is reproducible locally in trunk (5.1-SNAPSHOT), be sure to have empty
> data dir and execute:
> bin/cassandra -f
> then shut it down, start again. There is this in the logs:
> {code}
> INFO [main] 2025-04-03 16:36:04,572 CommitLog.java:206 - Replaying
> ./bin/../data/commitlog/CommitLog-9-1743690864518.log
> WARN [main] 2025-04-03 16:36:04,579 CommitLogReplayer.java:355 - Origin of 1
> sstables is unknown or doesn't match the local node; commitLogIntervals for
> them were ignored
> WARN [main] 2025-04-03 16:36:04,579 CommitLogReplayer.java:355 - Origin of 2
> sstables is unknown or doesn't match the local node; commitLogIntervals for
> them were ignored
> WARN [main] 2025-04-03 16:36:04,579 CommitLogReplayer.java:355 - Origin of 3
> sstables is unknown or doesn't match the local node; commitLogIntervals for
> them were ignored
> WARN [main] 2025-04-03 16:36:04,579 CommitLogReplayer.java:355 - Origin of 2
> sstables is unknown or doesn't match the local node; commitLogIntervals for
> them were ignored
> INFO [main] 2025-04-03 16:36:04,600 CommitLogReader.java:257 - Finished
> reading ./bin/../data/commitlog/CommitLog-9-1743690864518.log
> {code}
> There is no reason to have unknown origin, it is the only node in the cluster
> etc ... Pretty suspicious.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@cassandra.apache.org
For additional commands, e-mail: commits-h...@cassandra.apache.org
