This message was deleted.

Just curious, do you put the node into maintenance mode[1] before powering it down? [1] https://docs.harvesterhci.io/v1.1/host/#node-maintenance

Hi @square-crayon-51414, could you describe in more detail? This means how you disable the node. Did you cordon node as @bored-painting-68221 mentioned? or did you set up the no-schedule taint? Also, you saw the replica rebuild is processing, is the rebuild target a schedulable node?

thanks for hints. why i ask a guidance is i don't know the correct operation process. i tried kubectl drain node, with ignore-daemonsets flag and delete-emptydir-data flag, also tried cordon and maintain node via harvester ui, also tried these operation combination. i dont know how to verify if the node could be reboot safely. @salmon-city-57654, i am not sure which operation is setting noschedule taint, kubectl drain or harvester ui maintenance. Replicas rebuilding happen on the nodes which have ssd and scheduleable flag on, never reboot. i know the situation i'm facing might be my miss configuration, not harvester or longhorn issue. So i just ask a documents hints. thanks again.

Hi @square-crayon-51414 You would like to shut down some machines with volume replicas. Do I understand right? If yes, once you shut down any of these machines, that will trigger the longhorn rebuild. Maybe that is why you see the rebuild. To prevent the rebuild, you need to make all nodes are cordoned. The drain command would remove the replica, which also triggers a rebuild. Did you want to keep the whole replicas healthy when you shut down some nodes? As Connor mentioned, the document may introduce how to remove a node from the harvester cluster. That may not be very suitable for you. So I would like to realize your scenario more. Thanks!

Dear @salmon-city-57654, my scenario is power off 2 node every night and power on nextday morning. Removing node and joining the cluster may not suitable for me, because I have to do some node configuration, like storage network define, set longhorn inscheduleable. Your hints are greate helpful to me, like cordon all node to prevent rebuilding. I will do some testing. I want to stress one point that there isn't any replica on the node i powered off because i set inscheduleable flag on this node. It is alos why i'm confused power off these nodes tigger the rebuilding.