[
https://issues.apache.org/jira/browse/SLING-9118?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17082925#comment-17082925
]
Robert Munteanu commented on SLING-9118:
----------------------------------------
This is still on my to-do and the root cause is understood. I just need a bit
more time to find the best approach for solving it.
The cause (similar to SLING-7811) is the repository service restarting due to
the {{SecurityProviderRegistration}} unregistering and registering a new
{{SecurityProvider}} service. This in turn stops the repository service and
starts it again. Before SLING-7811 it failed with an NPE. After the (interim)
solution from SLING-7811 it uses {{Thread#interrupt}} which is usually a good
idea by strongly recommended against by Oak, see
https://jackrabbit.apache.org/oak/docs/dos_and_donts.html . I suspect this is
the root cause of the problems and am looking for a solution which allows the
repository to be shut down during startup in a timely manner and without using
interrupts.
> Sling fails to start when database exists but 'sling' directory is missing
> --------------------------------------------------------------------------
>
> Key: SLING-9118
> URL: https://issues.apache.org/jira/browse/SLING-9118
> Project: Sling
> Issue Type: Bug
> Reporter: Ben Radey
> Assignee: Robert Munteanu
> Priority: Major
> Attachments: drop-mongo.sh, error.log, recreateSlingReplicaSet.sh,
> run-mongo.sh, run-sling-initial.sh, run-sling-second.sh,
> sling-startup-error.log.txt
>
>
> # Create a persistent mongodb to use with Sling.
> # Start sling using the mongodb.
> # Stop sling.
> # Remove 'sling' directory.
> # Attempt to restart sling. Ultimately, sling fails to start.
> ----
> Steps to reproduce with attached scripts:
> # Run [^run-mongo.sh] . This creates a MongoDB 3.6 container named
> _mongo-sling_
> # Run [^run-sling-initial.sh]. This starts up Sling in the oak_mongo runmode,
> shuts it down after it's (probably) started up
> # Run [^run-sling-second.sh]. This moves away the sling directory and starts
> up a new instance. This always fails
> For cleanup, the [^drop-mongo.sh] script stops and removes the _mongo-sling_
> container.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
