Stopping Solr leaves .lock files behind

[joostmeijles]

When I stop my Solr container (using ltsc2019 image & process isolation) the write.lock files are not removed.
This causes a restart of the Solr container to fail.

Is this a known bug?

[pbering]

Haven't seen it on 9.0.2 Solr 1803 which is in production. Tested the ltsc2019 variant on Windows 10 using hyperv isolation and didn't see that issues. Can you please try if you can reproduce it using hyperv?

[joostmeijles]

Tested; with hyper-v on Windows 10 having (intermittently) the same issue.

[pbering]

Just tested on Windows 10 1809 with the 9.0.2 1803 solr with no problems, restarted containers multiple times. I can see that there is a write.lock in some cores but they don't seem to give Solr any issues starting up. Can you try to write step by step what you are doing so I can try the same?

[joostmeijles]

Steps are quite straightforward:

1. Start Solr container
2. Wait a bit for Solr to start / attach all cores
3. Stop Solr container
4. (Re-) start Solr container with same cores
5. Solr reports locking errors for the core directories that contain write.lock files

I have reproduced this intermittently at multiple machines with process- and Hyper-V isolation.
Although using process-isolation seems to trigger this problem more often.

I suspect that, when stopping a Solr container, depending on how quick it ends all processes, the cores correctly stopped.

My current plan is to simply remove any write.lock files upon start of the container.

[pbering]

in step 3 are you using stop or down

but anyways, it's a good plan to just wipe all write.lock files during startup, could be done in Boot.cmd

[joostmeijles]

Using stop in step 3.

[pbering]

was able to reproduce, so I fixed it. Thank you 👍