[03:38:00] <TimStarling>	 Multi-DC stage 1.5 is complete
[03:38:17] <TimStarling>	 it is time to schedule stage 2
[03:42:00] <TimStarling>	 I can announce it for early next week if there are no objections
[03:43:03] <TimStarling>	 in the meantime, I might have a go at the cluster maintenance problem, since we don't want to make mediawiki.org slow for everyone
[08:20:20] <Krinkle>	 TimStarling: ack. Perhaps we can reduce writing/populating db host side check to write attempts only. So passive checks on GET only consider configured mode and eg remembered state from an attempted write. That would presumably be effectively the same as today in terms of what we can observe on a busy enough wiki, depending on how long it's stored and where. Might be worthy moving to memc in that case. Or perhaps keep as today but move 
[08:20:20] <Krinkle>	 getWithSet to post send so we're at most one req behind per appserver