0

We've observed that a write.lock file can remain for Solr when the JVM falls over unexpectedly. This leaves the index locked despite the fact that nothing owns that lock.

Detailed here: Caused by: org.apache.solr.common.SolrException: Index locked for write for core

We intend to investigate switching to a SimpleFSLockFactory which is recommended for NFS environments, however, I would like to know what are the consequences of performing the manual action of deleting the write.lock file for Solr. Is there any possibility of data loss or corruption?

Providing of coarse that we ensure nothing owns the lock at that moment.

Community
  • 1
  • 1
kona
  • 1

1 Answers1

0

The write.lock is just an indicator that some Solr/JVM is using the core, so deleting it should not cause any more or less corruption than already existed as a result of the JVM failure. Assuming, as you point out, the core is actually not being used.

Using NFS for a solr index is definitely not considered a best practice though. Any read-path access that needs the NFS link will show terrible performance.

randomstatistic
  • 800
  • 5
  • 11