ignite-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Anton Vinogradov (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (IGNITE-2801) Coordinator floods network with partitions full map exchange messages
Date Mon, 21 Mar 2016 09:29:25 GMT

    [ https://issues.apache.org/jira/browse/IGNITE-2801?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15203937#comment-15203937
] 

Anton Vinogradov commented on IGNITE-2801:
------------------------------------------

Reproduced. 

Started 10 nodes with 35 caches. 
Gained 37 MB of GridDhtPartitionsFullMessage per minute -> 49 Mb/s
Rrough multiplication to 30 nodes & 65 caches (x6) gives exactly 300Mb/s 

> Coordinator floods network with partitions full map exchange messages
> ---------------------------------------------------------------------
>
>                 Key: IGNITE-2801
>                 URL: https://issues.apache.org/jira/browse/IGNITE-2801
>             Project: Ignite
>          Issue Type: Bug
>          Components: cache
>    Affects Versions: 1.5.0.final
>            Reporter: Denis Magda
>            Assignee: Anton Vinogradov
>            Priority: Critical
>              Labels: community, important
>             Fix For: 1.6
>
>         Attachments: basic_node.nps, basic_node.png, coordinator.nps, coordinator.png
>
>
> It is detected that the more machines in the cluster we have and the more caches are
started then the more outgoing traffic is produced by a coordinator node.
> As an example in the current deployment
> - 30 nodes;
> - 67 caches;
> - caches are empty and the cluster is not used at all (idle).
> the coordinator constantly uses 300 Mbit/s of outgoing traffic. In contrast each other
node shows constant 10 Mbit/s usage of incoming traffic.
> Most likely the reason is that the coordinator indefinitely sends partitions full map
for all the caches to all the nodes. This shouldn't happen.
> Need to debug the reason of the issue and fix it.
> Attached snapshots taken from the coordinator and on of cluster's nodes. Probably they
would help.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message