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

Posted to notifications@couchdb.apache.org by "Robert Newson (JIRA)" <ji...@apache.org> on 2017/03/01 10:50:46 UTC

[jira] [Commented] (COUCHDB-3311) Restore jitter-ed sleep to replicator manager shard discovery

    [ https://issues.apache.org/jira/browse/COUCHDB-3311?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15889959#comment-15889959 ] 

Robert Newson commented on COUCHDB-3311:
----------------------------------------

Context;

jitter(N) ->
    Range = min(2 * N * ?AVG_DELAY_MSEC, ?MAX_DELAY_MSEC),
    random:uniform(Range).

So, for N > 300, all jobs get a random jitter of MAX_DELAY_MSEC (60000).

I don't think this is adequate flow control, and we're also spawning processes we didn't make before, isn't that exacerbating the problem?



> Restore jitter-ed sleep to replicator manager shard discovery
> -------------------------------------------------------------
>
>                 Key: COUCHDB-3311
>                 URL: https://issues.apache.org/jira/browse/COUCHDB-3311
>             Project: CouchDB
>          Issue Type: Bug
>            Reporter: Nick Vatamaniuc
>
> Jitter-ed sleep or was dropped inadvertently when we switch to using mem3 shards instead of scanning the file system.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)