From discussion in #wikimedia-releng yesterday:
09:26:56 <legoktm> have we done failovers for Gerrit/Phab during the datacenter switchover in the past? ... 09:32:19 <hashar> Phabricator I can't tell, twentyafterfour will know whether it got switched over any recently 09:33:20 <twentyafterfour> we have a backup machine and we have done failover before but not as part of the datacenter switch as far as I can remember 09:34:19 <twentyafterfour> it would probably be good to exercise the process but it's not hard to failover phabricator 09:34:26 <twentyafterfour> mainly just a dns switch I think
We should:
- Document the process of failing over Phabricator, maybe on a wiki page
- Automate said process by implementing it in a cookbook
- Test regularly by including it in the normal Datacenter-Switchover
T164810: Switch phabricator production to codfw is related and might be helpful with the documentation step.