Issue Summary

If one reindex job is attempted to start before the previous job has released the index locks, it can cause the second reindex job to fail.

Steps to Reproduce

The admin UI thread must be opening a new IndexWriter on the issues directory at the exact moment the re-index thread is trying to close the old one and clean the directory. The new writer grabs the write.lock, and when the DefaultIndexEngine.clean() tries to create its writer to wipe the index, it fails because the lock is already held within the same JVM.

Expected Results

The new reindex job will acquire the lock and cancel the previous job.

Actual Results

The first reindex job does not release the lock and the second reindex job fails.

Workaround

Currently there is no known workaround for this behavior. A workaround will be added here when available