[
https://issues.apache.org/jira/browse/HBASE-21216?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16628994#comment-16628994
]
Ted Yu commented on HBASE-21216:
--------------------------------
The cause for the assertion failure was that the trigger of refresh for
SnapshotFileCache is indeterministic w.r.t. the timing of the assertion.
If cache of SnapshotFileCache is refreshed right before the collection of
archive files, archives may be cleaned by the SnapshotHFileCleaner.
I have a patch using which I loop the test locally.
So far 12 iterations have passed successfully.
Will upload the patch in a moment.
> TestSnapshotFromMaster#testSnapshotHFileArchiving is flaky
> ----------------------------------------------------------
>
> Key: HBASE-21216
> URL: https://issues.apache.org/jira/browse/HBASE-21216
> Project: HBase
> Issue Type: Test
> Reporter: Ted Yu
> Priority: Major
>
> From
> https://builds.apache.org/job/HBase-Flaky-Tests/job/branch-2/794/testReport/junit/org.apache.hadoop.hbase.master.cleaner/TestSnapshotFromMaster/testSnapshotHFileArchiving/
> :
> {code}
> java.lang.AssertionError: Archived hfiles [] and table hfiles
> [9ca09392705f425f9c916beedc10d63c] is missing snapshot
> file:6739a09747e54189a4112a6d8f37e894
> at
> org.apache.hadoop.hbase.master.cleaner.TestSnapshotFromMaster.testSnapshotHFileArchiving(TestSnapshotFromMaster.java:370)
> {code}
> The file appeared in archive dir before hfile cleaners were run:
> {code}
> 2018-09-20 10:38:53,187 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-archive/
> 2018-09-20 10:38:53,188 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |----data/
> 2018-09-20 10:38:53,189 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-------default/
> 2018-09-20 10:38:53,190 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |----------test/
> 2018-09-20 10:38:53,191 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-------------1237d57b63a7bdf067a930441a02514a/
> 2018-09-20 10:38:53,192 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |----------------recovered.edits/
> 2018-09-20 10:38:53,193 DEBUG [Time-limited test] util.CommonFSUtils(774):
> |-------------------4.seqid
> 2018-09-20 10:38:53,193 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-------------29e1700e09b51223ad2f5811105a4d51/
> 2018-09-20 10:38:53,194 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |----------------fam/
> 2018-09-20 10:38:53,195 DEBUG [Time-limited test] util.CommonFSUtils(774):
> |-------------------2c66a18f6c1a4074b84ffbb3245268c4
> 2018-09-20 10:38:53,196 DEBUG [Time-limited test] util.CommonFSUtils(774):
> |-------------------45bb396c6a5e49629e45a4d56f1e9b14
> 2018-09-20 10:38:53,196 DEBUG [Time-limited test] util.CommonFSUtils(774):
> |-------------------6739a09747e54189a4112a6d8f37e894
> {code}
> However, the archive dir became empty after hfile cleaners were run:
> {code}
> 2018-09-20 10:38:53,312 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-archive/
> 2018-09-20 10:38:53,313 DEBUG [Time-limited test] util.CommonFSUtils(771):
> |-corrupt/
> {code}
> Leading to the assertion failure.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
