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

Posted to issues@ratis.apache.org by "Lokesh Jain (JIRA)" <ji...@apache.org> on 2019/05/01 07:03:00 UTC

[jira] [Commented] (RATIS-435) RaftLogWorker uses a blocking queue which blocks append entries

    [ https://issues.apache.org/jira/browse/RATIS-435?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16830893#comment-16830893 ] 

Lokesh Jain commented on RATIS-435:
-----------------------------------

We might accumulate too many blocked requests from leader in the follower. These requests might take up extra memory as well given that these can be chunk write requests in case of Ozone. Also by failing the leader knows that the follower is overburdened and leader can use it to throttle the requests to the follower.

In future it can also let the client know of possible slowness in a particular follower. In case of ozone it can be used to convey SCM of possibly slow pipelines.

> RaftLogWorker uses a blocking queue which blocks append entries
> ---------------------------------------------------------------
>
>                 Key: RATIS-435
>                 URL: https://issues.apache.org/jira/browse/RATIS-435
>             Project: Ratis
>          Issue Type: Bug
>            Reporter: Lokesh Jain
>            Assignee: Lokesh Jain
>            Priority: Major
>              Labels: ozone
>
> Currently RaftLogWorker uses a blocking queue which blocks the append entries if the queue is full. We should rather fail the operation if the queue is full.



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