Comment 33 for bug 240437

OK, sorry for the delay Roger. So if I understand correctly, the only way to reproduce this problem is to create a new user when the users list appears empty? That would be logical, since we commit the data we have, which would be empty if it could not be loaded.

But do you mean that the second time you start users-admin, the users list is empty even if you don't get the error message? That would be the real failure here: we should never allow the tool to run if retrieving data failed.

Then, another part of the issue is that the backends seem to be too slow to start before the timeout is reached. That may come from the fact perl needs quite a bit of memory to run, which may need some swapping. That's relatively hard to solve, but at the very least we should display an error, not destroy the system's configuration.

Xande: are you sure users-admin is starts faster after restoring the backup? That seems really strange if the files are the same. One interesting test (for both of you) would be to start the scripts before starting users-admin, and to compare the startup times: start users-admin once, run it again and see the time it needs, then wait for 5 minutes so that the scripts to stop automatically, and restart users-admin to compare. The first time is needed so that the tool is loaded in memory, but it should not be taken as reference for anything.