You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@uima.apache.org by "Lou DeGenaro (JIRA)" <de...@uima.apache.org> on 2019/01/03 13:45:00 UTC

[jira] [Commented] (UIMA-5928) DUCC Agent quiesce should wait (forever) for non-fairshare displatchables to complete before shutting down

    [ https://issues.apache.org/jira/browse/UIMA-5928?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16733044#comment-16733044 ] 

Lou DeGenaro commented on UIMA-5928:
------------------------------------

RM changes to handle quiesced Node:

> IRmPersistence

     - add new enum RmNodes.Quiesced

> ResourceManagerEventListener

     - add onNodeMetricsEvent method debugging to log NodeStatus comprising "UnAvailable" (meaning quiesced).

> NodePool

     - add new List<Node> quiesceMachines comprising those nodes that have been quiesced

     - add to list and update DB when node reports "quiesced" and not already on list

     - remove from list and update DB when node reports not "quiesced" and in list

     - employ list when determining if node is schedulable

WS changes to handle quiesced Node:

> update system.daemons.jsp to include new column Quiesce

> update processing of request for system machines page to fill-in T/F for quiesce status (classic and scroll) as recorded by RM to DB

> DUCC Agent quiesce should wait (forever) for non-fairshare displatchables to complete before shutting down
> ----------------------------------------------------------------------------------------------------------
>
>                 Key: UIMA-5928
>                 URL: https://issues.apache.org/jira/browse/UIMA-5928
>             Project: UIMA
>          Issue Type: Improvement
>          Components: DUCC
>            Reporter: Lou DeGenaro
>            Assignee: Jerry Cwiklik
>            Priority: Major
>             Fix For: 2.2.3-Ducc
>
>
> {color:#121212}When stop_ducc -c agent@node --quiesce-then-stop is issued, the agent should immediately stop any managed PIDs running in a pre-emptable class, but wait for all managed PIDs running in a non-preemptable class to be terminated before stopping.{color}



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