Single-node instance (for testing) sometimes already has a database

Using FoundationDB
1

I have a few tests that bring up fdbserver and then try to configure it using configure new single memory.

This succeeds most of the time, but sometimes fails because database already exists (the error suggests not using new), and configure single memory works.

I purge the directory every time the test finishes, so I don’t expect it to have any side-effects across runs. I did double-check to make sure I wasn’t talking to the fdbserver process I brought up for the previous run. So the question is:

  • Is a newly started single-node fdbserver -l public -p 127.0.0.1:4500 process sometimes supposed to have a database? (so is configure new... supposed to fail)?
  • Is there something I have overlooked which leads to this state in my newly started process? Everything including cluster-file is in the directory that I purge after every run, successful or not.

Because this is a test, I run configure single memory when configure new single memory fails and this works reliably, but this hack doesn’t feel right.

2

Also, this comment appears to suggest that configure may be necessary: Error: The database is unavailable - #3 by ajbeamon