[
https://issues.apache.org/jira/browse/CONNECTORS-880?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13891073#comment-13891073
]
Karl Wright commented on CONNECTORS-880:
----------------------------------------
The problem is happening during the cleanup phase of the crawl, most likely
because of an error from the output connector.
Normally:
JobResetThread calls JobManager.resetJobs() which normally calls
Jobs.finishJob(time).
Error condition:
DocumentCleanupThread calls JobManager.errorAbort(), which is clearly the wrong
method to use, since it just throws an exception for any jobs in the
STATUS_SHUTTINGDOWN state: throw new ManifoldCFException("Job "+jobID+"
is not active");
So, the SHUTTINGDOWN state needs its own abort method, e.g. cleanupAbort(), and
possibly its own state (ABORTING_CLEANINGUP).
> Under the right conditions, job aborts do not update "last checked" time
> ------------------------------------------------------------------------
>
> Key: CONNECTORS-880
> URL: https://issues.apache.org/jira/browse/CONNECTORS-880
> Project: ManifoldCF
> Issue Type: Bug
> Components: Framework crawler agent
> Affects Versions: ManifoldCF 1.4.1
> Reporter: Karl Wright
> Assignee: Karl Wright
> Fix For: ManifoldCF 1.6
>
>
> When a scheduled job is being considered to be started, MCF updates the
> last-check field ONLY if the job didn't start. It relies on the job's
> completion to set the last-check field in the case where the job does start.
> But if the job aborts, in at least one case the last-check field is NOT
> updated. This leads to the job being run over and over again within the
> schedule window.
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)
