You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@ignite.apache.org by "Kirill Gusakov (Jira)" <ji...@apache.org> on 2022/12/30 14:07:00 UTC

[jira] [Comment Edited] (IGNITE-18447) Hanging in RebalanceRaftGroupEventsListener

    [ https://issues.apache.org/jira/browse/IGNITE-18447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17653141#comment-17653141 ] 

Kirill Gusakov edited comment on IGNITE-18447 at 12/30/22 2:06 PM:
-------------------------------------------------------------------

We already have the separate issue to remove sync gets in this peace of code https://issues.apache.org/jira/browse/IGNITE-17592. Also, these listeners will be reworked/removed during the current rebalance redesign https://issues.apache.org/jira/browse/IGNITE-18290. *So, sync calls definetly should be removed.*

At the same time, as I uderstand, at the moment this issue isn't prevent node from stop completely, sometimes we will wait for raft timeout in the worst case.


was (Author: kgusakov):
We already have the separate to remove sync gets in this peace of code https://issues.apache.org/jira/browse/IGNITE-17592. Also, these listeners will be reworked/removed during the current rebalance redesign https://issues.apache.org/jira/browse/IGNITE-18290. *So, sync calls definetly should be removed.*

At the same time, as I uderstand, at the moment this issue isn't prevent node from stop completely, sometimes we will wait for raft timeout in the worst case.

> Hanging in RebalanceRaftGroupEventsListener
> -------------------------------------------
>
>                 Key: IGNITE-18447
>                 URL: https://issues.apache.org/jira/browse/IGNITE-18447
>             Project: Ignite
>          Issue Type: Bug
>            Reporter: Andrey Mashenkov
>            Priority: Major
>              Labels: ignite-3
>
> RebalanceRaftGroupEventsListener schedules a task that waits for metastorage operation future under readlock (busy-lock).
> This prevents TableManager from being stopped because, TableManager waits for a writelock (busy-lock, which is reused in RebalanceRaftGroupEventsListener) before shutdown the scheduler thread.
> As of now, the issue can't be reproduced because any raft command are cancelled with TimeoutException eventually, and the scheduled task will ignore/log the exception.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)