[
https://issues.apache.org/jira/browse/IGNITE-8167?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16428409#comment-16428409
]
ASF GitHub Bot commented on IGNITE-8167:
----------------------------------------
GitHub user amelius0712 opened a pull request:
https://github.com/apache/ignite/pull/3771
IGNITE-8167: Fix inconsistent last record pointer in case of recovery from
corrupted WAL
Let's look at this peace of code from
GridCacheDatabaseSharedManager.readCheckpointAndRestoreMemory
`
WALPointer restore = restoreMemory(status);
// First, bring memory to the last consistent checkpoint state
if needed.
// This method should return a pointer to the last valid record
in the WAL.
cctx.wal().resumeLogging(restore);
`
In case of `restore == null`. Logging will be resuming from 0 absolute WAL
index.
You can merge this pull request into a Git repository by running:
$ git pull https://github.com/Synesis-LLC/ignite ignite-8167
Alternatively you can review and apply these changes as the patch at:
https://github.com/apache/ignite/pull/3771.patch
To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:
This closes #3771
----
commit 40b9c8e227783f8d90fff5f2db4688e63be3dd37
Author: Pavel Sapezhko <pavel.sapezhko@...>
Date: 2018-04-06T14:36:23Z
IGNITE-8167: Fix inconsistent last record pointer in case of recovery from
corrupted WAL
----
> Recovery after crash sometimes leads to starting from beginning absolute wal
> segment index
> ------------------------------------------------------------------------------------------
>
> Key: IGNITE-8167
> URL: https://issues.apache.org/jira/browse/IGNITE-8167
> Project: Ignite
> Issue Type: Bug
> Affects Versions: 2.4
> Environment: Doesn't meter. We saw these behavior in k8s deployment
> as in local deployment too. Using any of WAL_MOD.
> Reporter: Pavel Sapezhko
> Priority: Major
> Fix For: 2.5
>
>
> When we are trying to restore after crash using wal log, sometimes we can
> find corrupted wal messages which can leads to starting from beginning
> absolute wal index. So, we will have broken wal archiver thread due to
> assertation error(but we still having working Ignite instance. I think we
> need to discuss if we are really want it) and as a result on next restart we
> can see "Wal history is too short" message.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
