[
https://issues.apache.org/jira/browse/IGNITE-27685?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vladimir Pligin updated IGNITE-27685:
-------------------------------------
Component/s: rw transactions ai3
> Implement exponential backoff for durable finish
> ------------------------------------------------
>
> Key: IGNITE-27685
> URL: https://issues.apache.org/jira/browse/IGNITE-27685
> Project: Ignite
> Issue Type: Improvement
> Components: rw transactions ai3
> Reporter: Alexey Scherbakov
> Priority: Major
> Labels: ignite-3
>
> Currently "durable finish" [1] process is retried without any backoff for
> each transaction independently.
> This can cause CPU usage spikes in various unavailability scenarios, for
> example, until new primary replica is elected for commit partition.
> Also we need to inroduce new metrics related to durable finish process for
> better observability.
> [1] org.apache.ignite.internal.tx.impl.TxManagerImpl#durableFinish
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
