This message was deleted.
# longhorn-storagea
adamant-kite-43734
08/23/2023, 1:13 AMThis message was deleted.
b
blue-kitchen-51801
08/23/2023, 6:24 AMI'm also interested for an answer
n
narrow-egg-98197
08/23/2023, 7:05 AMThe approach I'm currently thinking of is indeed the approach you said, and it's similar to the approach mentioned in the document: Updating the Node OS or Container Runtime.
narrow-egg-98197
08/23/2023, 7:06 AMThe method that can speed up the rebuilding first is to adjust the
Replica Replenishment Wait Intervalnarrow-egg-98197
08/23/2023, 7:08 AMWhat version number are you currently using?
n
nice-tent-65195
08/23/2023, 1:40 PMwe are experiencing the same problem with 1.5.1 LH
in our case, we do not exceed the default Replica Replenishment Wait Interval (with the node drain state)
so increasing it would not help with the problem
a
abundant-hair-58573
08/23/2023, 2:20 PMWe are at 1.3.2. I actually reduced the replenishment wait interval so the new replicas would start building sooner, since I was waiting for that to finish before draining and rebooting the next node. Many of our nodes are pretty full and Unschedulable, if I get the node back up within the Replinishment wait interval will it still rebuild the replica on that same node even if it's unschedulable?
abundant-hair-58573
08/23/2023, 5:23 PMWith a little testing it looks like Longhorn will not rebuild the failed replica on the same node when it comes back up if it is Unschedulable. At least half of our nodes are unschedulable so that's unfortunate.
4 Views