So, we had an outage a couple days back. It would appear that seconds after a node was put in maintenance, a disk on another node failed. Am aware that writes/changes to the node in maintenance proceed to other nodes until maintenance is exited, but it got me wondering – if there is only two copies of the data, a disk fails, and the other copy of that data is in a node under maintenance… if that node is still up and available even tho it’s in maintenance, will it supply the data to rebuild the copy elsewhere to keep the volumes going, given the failed disk?
Or if a node is under maintenance, does that mean “strictly no writes, and also no reads even though I could potentially satisfy them until exited” ? I suspect it’s the latter because if the maintenance node is not taking part in changes to data, then how can it know the data it is supplying as read is current? Or is there wizardry happening that would say “well, in this exceptional circumstance, here is the data you need”?
Can anyone offer any clarity here? Thanks