[
https://issues.apache.org/jira/browse/IGNITE-18424?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksey Plekhanov updated IGNITE-18424:
---------------------------------------
Release Note: Fixed WAL timeout rollover
> WAL segment rollover does not happen after walForceArchiveTimeout
> -----------------------------------------------------------------
>
> Key: IGNITE-18424
> URL: https://issues.apache.org/jira/browse/IGNITE-18424
> Project: Ignite
> Issue Type: Bug
> Affects Versions: 2.14
> Reporter: Ilya Shishkov
> Assignee: Nikolay Izhikov
> Priority: Major
> Labels: IEP-59, ise
> Fix For: 2.15
>
> Time Spent: 20m
> Remaining Estimate: 0h
>
> How to reproduce:
> # Start and activate persistent cluster with such properties:
> ## {{DataStorageConfiguration#walForceArchiveTimeout > 0}}, eg. 10000 millis.
> ## {{DataRegion#cdcEnabled=true}}.
> ## {{DataStorageConfiguration#walSegmentSize=1Gb}}.
> ## {{DataStorageConfiguration#maxWalArchiveSize=20Gb}}.
> # Start data changing (put some entries continuously).
> # Rollover of WAL is not happening. You can wait some period exceeding
> {{walForceArchiveTimeout}} to ensure.
> # Stop loading the data.
> # Rollover happens.
> If you does not stop to load the data, rollover _will not happen until WAL
> segment will beŃome full, i.e until to rollover by a segment size_.
> Under low loads, filling of 1Gb segment can occur after a dozen of minutes or
> even after several hours, which causes a large lag of a CDC replication.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
