Comment 12 for bug 1491576

We applied workaround which will minimise occurrence to almost 0

Firstly, logrotate on controllers will be a bit shuffled by adding random delay to every controller. IT won't run at the same particular time on all controllers.

However, when apache with keystone dies it will have side effect for a couple of connections/tokens as there is a bit delay between
apache2clt graceful || apache2ctl restart as apache will lose all connections. Also it requires time to start all workers.

I am assigning this bug to 8.0 to keystone team for further investigation of https://github.com/GrahamDumpleton/mod_wsgi/issues/81