[
https://issues.apache.org/jira/browse/HBASE-21387?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16675857#comment-16675857
]
Josh Elser commented on HBASE-21387:
------------------------------------
{quote}
Please take a look at two-pass-cleaner.v4.txt where cleaner chore keeps track
of the files deemed cleanable from previous iteration.
Only files deemed cleanable from previous and current iterations would be
deleted.
{quote}
I think that would help, but I'm not sure if that's the best way to go about
it. We're holding onto file names in memory (which could get big) with your
two-pass-cleaner.v4 patch. If we made a change to not clean files while there
is an in-progress snapshot, we would be in trouble if we ever got into a
situation where a snapshot was "orphaned" and prevent file cleaning from
happening.
I'm not sure which I think is better (or if there's even still something better
out there..)
> Race condition surrounding in progress snapshot handling in snapshot cache
> leads to loss of snapshot files
> ----------------------------------------------------------------------------------------------------------
>
> Key: HBASE-21387
> URL: https://issues.apache.org/jira/browse/HBASE-21387
> Project: HBase
> Issue Type: Bug
> Reporter: Ted Yu
> Assignee: Ted Yu
> Priority: Major
> Labels: snapshot
> Attachments: 21387.dbg.txt, 21387.v2.txt, 21387.v3.txt, 21387.v6.txt,
> 21387.v7.txt, 21387.v8.txt, two-pass-cleaner.v4.txt
>
>
> During recent report from customer where ExportSnapshot failed:
> {code}
> 2018-10-09 18:54:32,559 ERROR [VerifySnapshot-pool1-t2]
> snapshot.SnapshotReferenceUtil: Can't find hfile:
> 44f6c3c646e84de6a63fe30da4fcb3aa in the real
> (hdfs://in.com:8020/apps/hbase/data/data/.../a/44f6c3c646e84de6a63fe30da4fcb3aa)
> or archive
> (hdfs://in.com:8020/apps/hbase/data/archive/data/.../a/44f6c3c646e84de6a63fe30da4fcb3aa)
> directory for the primary table.
> {code}
> We found the following in log:
> {code}
> 2018-10-09 18:54:23,675 DEBUG
> [00:16000.activeMasterManager-HFileCleaner.large-1539035367427]
> cleaner.HFileCleaner: Removing:
> hdfs:///apps/hbase/data/archive/data/.../a/44f6c3c646e84de6a63fe30da4fcb3aa
> from archive
> {code}
> The root cause is race condition surrounding in progress snapshot(s) handling
> between refreshCache() and getUnreferencedFiles().
> There are two callers of refreshCache: one from RefreshCacheTask#run and the
> other from SnapshotHFileCleaner.
> Let's look at the code of refreshCache:
> {code}
> if (!name.equals(SnapshotDescriptionUtils.SNAPSHOT_TMP_DIR_NAME)) {
> {code}
> whose intention is to exclude in progress snapshot(s).
> Suppose when the RefreshCacheTask runs refreshCache, there is some in
> progress snapshot (about to finish).
> When SnapshotHFileCleaner calls getUnreferencedFiles(), it sees that
> lastModifiedTime is up to date. So cleaner proceeds to check in progress
> snapshot(s). However, the snapshot has completed by that time, resulting in
> some file(s) deemed unreferenced.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
