[
https://issues.apache.org/jira/browse/NIFI-4322?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16162892#comment-16162892
]
Mark Payne commented on NIFI-4322:
----------------------------------
This issue appears to manifest itself in a few ways. First, the FlowFile
repository can grow unbounded, taking up a huge amount of disk space. This can
exhaust the available space but if it does not, it can also lead to taking a
huge amount of time for NiFi to restart. I've seen repositories grow to
hundreds of GB, which took hours to recover from.
Secondly, because the FlowFile Repo is not checkpointing, it does not allow the
content repository to destroy or archive Content Claims that are no longer in
use. As a result, this can lead to the Content Repository growing unbounded
until it runs out of disk space.
> If RuntimeException thrown during checkpoint of FlowFile Repository, repo
> stops checkpointing
> ---------------------------------------------------------------------------------------------
>
> Key: NIFI-4322
> URL: https://issues.apache.org/jira/browse/NIFI-4322
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Mark Payne
> Assignee: Mark Payne
>
> The WriteAheadFlowFileRepository has a 'checkpointRunnable' task that is
> scheduled periodically, which is responsible for checkpointing the FlowFile
> Repository. It catches IOException but no other Exceptions. So if an OOME
> occurs or a RuntimeException, the repository will stop checkpointing. As a
> result, it can fill all disk space and/or take a very long time to recover
> upon restart, depending on how large the FlowFile repository has become.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
