You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@hbase.apache.org by "Duo Zhang (JIRA)" <ji...@apache.org> on 2018/09/27 09:49:00 UTC

[jira] [Commented] (HBASE-21233) Allow the procedure implementation to skip persistence of the state after a execution

    [ https://issues.apache.org/jira/browse/HBASE-21233?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16630078#comment-16630078 ] 

Duo Zhang commented on HBASE-21233:
-----------------------------------

Introduce a 'persist' flag in Procedure to indicate whether we need to persist the procedure after exection. Default to true. The implementation can set it to false to avoid persisting. And before execution, we will reset it to true every time.

Notice that, this can reduce the number procedure wals but can not fix all the problem for retrying. For TRSP it is OK, as the holdLock is true, but for other procedures where holdLock is false, after the execution we have to release the lock and persist the 'release lock' action, so skipPersistence here can reduce one procedure wal record, but the 'release lock' one is still needed.

This is just a framework for skip persistence. Will open other issues to fix the retrying for specific procedures.

> Allow the procedure implementation to skip persistence of the state after a execution
> -------------------------------------------------------------------------------------
>
>                 Key: HBASE-21233
>                 URL: https://issues.apache.org/jira/browse/HBASE-21233
>             Project: HBase
>          Issue Type: Sub-task
>          Components: Performance, proc-v2
>            Reporter: Duo Zhang
>            Assignee: Duo Zhang
>            Priority: Major
>             Fix For: 3.0.0, 2.2.0, 2.1.1
>
>         Attachments: HBASE-21233.patch
>
>
> Discussed with [~stack] and [~allan163] on HBASE-21035, that when retrying we do not need to persist the procedure state every time, as the retry timeout is not a critical stuff. It is OK that we loss this information and start from 0 when after restarting.



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