[
https://issues.apache.org/jira/browse/NIFI-1696?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Joseph Witt updated NIFI-1696:
------------------------------
Description:
The scenario:
- 3 node cluster
- node 'NodeA' is primary and running a processor 'ProcP1' set to event driven
scheduling
- 'ProcP1' wasn't working due to a config issue but there were a few events on
its input queues.
- restarted node 'NodeA' after fixing config issue.
- No new events had come in but those events were sitting there.
Result: The 'ProcP1' appears to be in scheduled state but does not get
scheduled to run - perhaps because no new event triggered it? Seems like
perhaps we're not checking if there are already events at restart? Stopping
and starting 'ProcP1' then resulted in things going back to normal.
It is not obvious to me how to recreate this in a repeatable way.
was:
The scenario:
- 3 node cluster
- node 'NodeA' is primary and running a processor 'ProcP1' set to event driven
scheduling on primary node only.
- restarted node 'NodeA'
Result: The 'ProcP1' appears to be in scheduled state but does not get
scheduled to run. Stopping and starting 'ProcP1' then resulted in things going
back to normal. Seems like a logic flaw must exist for such a case. Will try
to recreate on another cluster to be sure.
> Primary node event driven tasks appear not to auto restart
> ----------------------------------------------------------
>
> Key: NIFI-1696
> URL: https://issues.apache.org/jira/browse/NIFI-1696
> Project: Apache NiFi
> Issue Type: Bug
> Affects Versions: 0.6.0
> Reporter: Joseph Witt
> Priority: Critical
> Fix For: 0.7.0
>
>
> The scenario:
> - 3 node cluster
> - node 'NodeA' is primary and running a processor 'ProcP1' set to event
> driven scheduling
> - 'ProcP1' wasn't working due to a config issue but there were a few events
> on its input queues.
> - restarted node 'NodeA' after fixing config issue.
> - No new events had come in but those events were sitting there.
> Result: The 'ProcP1' appears to be in scheduled state but does not get
> scheduled to run - perhaps because no new event triggered it? Seems like
> perhaps we're not checking if there are already events at restart? Stopping
> and starting 'ProcP1' then resulted in things going back to normal.
> It is not obvious to me how to recreate this in a repeatable way.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
