Hello, I wonder if anyone could help me out here with some approaches to troubleshooting an issue I am encountering.
I got Neo4j community edition running in a container in a k8s environment (OpenShift Container Platform).
We encountered an issue at a customer site, whereby Neo4j is failing to start. Full log attached. This extract points to issue reading the transaction database:
'org.neo4j.dbms.database.DefaultSystemGraphInitializer@d5b599d6' was successfully initialized, but failed to start. Please see the attached cause exception "Unable to read log version and last committed tx from '/data/transactions/system/neostore.transaction.db.0'. Was only able to read -1 bytes, but was expecting 8".
If we look on disk, we can see that the file /data/transactions/system/neostore.transaction.db.0 does exist, is readable, but it is 0 size.
This problem is only evident at one particular customer site. There are many hundreds of successful installations in house and at other customer sites.
If the pod that contains the neo4j container is started without persistence configured, the problem goes away. When persistence is configured, the problem comes back. The PV and PVC are successfully created and PVC is being written to by neo4j - so it doesn't look like a storage permissions problem - but that database transaction file is always 0 length.
We have been unable to reproduce the issue using the same storage type in house (IBM Spectrum Scale).
Any thoughts on this? Any reasons why the transactions file would have 0 size on startup? Any known possibly related issues?
Thanks a lot.
