You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@sling.apache.org by "Christian Schneider (Jira)" <ji...@apache.org> on 2020/07/10 12:04:00 UTC

[jira] [Comment Edited] (SLING-9577) Switch back to seeding thread

    [ https://issues.apache.org/jira/browse/SLING-9577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17155397#comment-17155397 ] 

Christian Schneider edited comment on SLING-9577 at 7/10/20, 12:03 PM:
-----------------------------------------------------------------------

{quote}The same situation could happen with the current code when the single message sent out does not arrive for some reason.
{quote}
>No because it's an assign request and we assign to an offset that we know correspond to a >message created before the new seed message.

Yes .. but in case of an overload the consume might not work while the send works. So our queues are still stuck as they are not considered being seeded.


was (Author: marett):
bq. The same situation could happen with the current code when the single message sent out does not arrive for some reason.

No because it's an assign request and we assign to an offset that we know correspond to a message created before the new seed message.

> Switch back to seeding thread
> -----------------------------
>
>                 Key: SLING-9577
>                 URL: https://issues.apache.org/jira/browse/SLING-9577
>             Project: Sling
>          Issue Type: Improvement
>          Components: Content Distribution
>    Affects Versions: Content Distribution Journal Core 0.1.16
>            Reporter: Christian Schneider
>            Assignee: Christian Schneider
>            Priority: Major
>             Fix For: Content Distribution Journal Core 0.1.18
>
>
> The current code uses a combination of seeding thread, persisting and loading offsets from sling repo and sending single seeding messages.
> In sum this means we send at least one seeding message (seeding thread) on first run and one seeding message on following runs.
> I propose to switch back to a pure seeding thread solution and make sure it terminates correctly. This solution should in almost all cases also just send 1 message and is a lot simpler.
>  



--
This message was sent by Atlassian Jira
(v8.3.4#803005)