I noticed that when the NFS snapshot repo is unavailable during the elasticsearch service startup (e.g. is not mounted) the service refuses to start.
Is this the expected behavior?
I'd guess that issuing a warning and refusing the create a snapshot (obviously) would be better than refusing to start up at all. I mean, even if you ensure that the repo is available at startup it might become unavailable while the service is running and creating a snapshot would fail again...
The expected behavior is to create an empty directory for the snapshot path on the shared filesystem if it does not exist (the snapshot directory has to be a descendant of one of the path.repo). If the path is not accessible (mounted on a device which is unavailable) it will fail. This is a simple check for the validity of the repo. Maybe add the repository after the ES started and the NFS has been successfully mounted?
If I am missing something, please post the stack trace
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.