[
https://issues.apache.org/jira/browse/IGNITE-28079?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Roman Puchkovskiy updated IGNITE-28079:
---------------------------------------
Description: Currently, if disk I/O is saturated, huge puts (100Mb of size)
will not be able to be written to the log on the leader in time, so one of
ignite.raft.retryTimeoutMillis and ignite.replication.rpcTimeoutMillis will
cause a timeout. They can be configured to higher values (as a workaround), but
it seems that we should do some kinds of retries internally to create
backpressure here.
> Backpressure mechanism to account for slow partition Raft
> ---------------------------------------------------------
>
> Key: IGNITE-28079
> URL: https://issues.apache.org/jira/browse/IGNITE-28079
> Project: Ignite
> Issue Type: Improvement
> Reporter: Roman Puchkovskiy
> Priority: Major
> Labels: ignite-3
>
> Currently, if disk I/O is saturated, huge puts (100Mb of size) will not be
> able to be written to the log on the leader in time, so one of
> ignite.raft.retryTimeoutMillis and ignite.replication.rpcTimeoutMillis will
> cause a timeout. They can be configured to higher values (as a workaround),
> but it seems that we should do some kinds of retries internally to create
> backpressure here.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
