[
https://issues.apache.org/jira/browse/ACCUMULO-3838?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14553134#comment-14553134
]
Josh Elser commented on ACCUMULO-3838:
--------------------------------------
Re-reading the master log, it is possible that the test case just didn't read
the metadata table before the master deleted the status record from metadata
(after it copied it to the replication table).
{noformat}
2015-05-20 07:33:03,043 [replication.StatusMaker] DEBUG: Creating replication
status record for
file:/accumulo/test/target/mini-tests/org.apache.accumulo.test.replication.ReplicationIT_replicationEntriesPrecludeWalDeletion/accumulo/wal/hostname+59851/5dfa3257-859c-4ecd-9b4f-4abe5bbd04ac
on table 1 with [begin: 0 end: 0 infiniteEnd: true closed: true createdTime:
1432107178863].
2015-05-20 07:33:03,046 [replication.StatusMaker] INFO : Creating order record
for
file:/accumulo/test/target/mini-tests/org.apache.accumulo.test.replication.ReplicationIT_replicationEntriesPrecludeWalDeletion/accumulo/wal/hostname+59851/5dfa3257-859c-4ecd-9b4f-4abe5bbd04ac
for 1 with [begin: 0 end: 0 infiniteEnd: true closed: true createdTime:
1432107178863]
2015-05-20 07:33:03,048 [replication.StatusMaker] DEBUG: Deleting
~replfile:/accumulo/test/target/mini-tests/org.apache.accumulo.test.replication.ReplicationIT_replicationEntriesPrecludeWalDeletion/accumulo/wal/hostname+59851/5dfa3257-859c-4ecd-9b4f-4abe5bbd04ac
stat:1 [] 5 false from metadata table as it's no longer needed
{noformat}
If the Master saw the status record in accumulo.metadata, copied it to
replication table, deleted it from metadata before the test's thread could read
the status message from metadata, the test could fail as described above.
> ReplicationIT.replicationEntriesPrecludeWalDeletion failed because it missed
> an expected WAL
> --------------------------------------------------------------------------------------------
>
> Key: ACCUMULO-3838
> URL: https://issues.apache.org/jira/browse/ACCUMULO-3838
> Project: Accumulo
> Issue Type: Bug
> Components: test
> Affects Versions: 1.7.0
> Reporter: Josh Elser
> Assignee: Josh Elser
> Priority: Minor
> Fix For: 1.8.0, 1.7.1
>
>
> Automated testing found a situation where we have 4 WALs in the replication
> table, but only expected to find 3. Thus, the test failed.
> Looking at the master log, there shouldn't be any way that the test _didn't_
> see this log. My hunch is that the test accidentally removed one of the logs
> that should have been left in the expected set.
> Need to revisit the logic to make sure it's solid.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
