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

Posted to issues@ignite.apache.org by "Dmitriy Pavlov (JIRA)" <ji...@apache.org> on 2018/03/12 15:06:00 UTC

[jira] [Commented] (IGNITE-7920) Research local crash recovery speed after stop in the middle of checkpoint

    [ https://issues.apache.org/jira/browse/IGNITE-7920?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16395366#comment-16395366 ] 

Dmitriy Pavlov commented on IGNITE-7920:
----------------------------------------

It is required to check if WAL iterator uses underlying buffer to read entries.
It is also required to check if some information may be cached during restore, to avoid read twice.

> Research local crash recovery speed after stop in the middle of checkpoint
> --------------------------------------------------------------------------
>
>                 Key: IGNITE-7920
>                 URL: https://issues.apache.org/jira/browse/IGNITE-7920
>             Project: Ignite
>          Issue Type: Task
>          Components: persistence
>            Reporter: Dmitriy Pavlov
>            Assignee: Dmitriy Pavlov
>            Priority: Major
>
> When using Ignite as a database on one node, I observed that the recovery time using the WAL exceeds the time between checkpoints.
> Ignite logs message that it is restored after stopping in the middle of the checkpoint 
> I use the frequency of CP - setCheckpointFrequency (60 * 1000), but at server with HDD and heavy IO load. At least 16 minutes required to restore Ignite after stop.
> Server monitoring states that  high rate of IOPS ocurred and it is constraint (there is plenty of CPU/RAM)



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)