[07:16:55] <jynus>	 good morning
[07:20:02] <Emperor>	 hi
[07:55:17] <Emperor>	 bit concerned about the power work proposal - AIUI we have 3-way replication in ms-swift, which means that turning off 5 nodes at once (codfw A2 contains ms-be20{28,29,40,44,51}) means there could be availability issues. And even though eqiad is currently active DC (and we could depool codfw-swift for reads) MW will try and write to both swift clusters. Is that all correct, godog?
[07:57:18] <godog>	 Emperor: data is replicated across rows, one rack (or one row) shouldn't cause availability issues
[08:00:47] <Emperor>	 godog: Ah, I'd missed that detail; how is that set up? the rings not explicitly have row in them, do they? Or is that reflected in IP address somehow (and thus zone)?
[08:02:03] <godog>	 Emperor: yes that's correct, each row has its own subnet and thus reflected in the ring as zones
[08:23:14] <Emperor>	 TIL, thank you :)
[08:28:37] <godog>	 for sure!
[08:52:17] <Emperor>	 that does mean we'll want to make sure e.g. all the A7 servers are back up and happy before they start on the row B PDUs (some of which are also due the same day). I'll write something up for Papaul