maintenance practices.... Distributed migrateall?

wolfspyre

Member
Jan 6, 2023
23
4
8
48
Austin, TX
wolfspyre.com
Howdy all!

I'm moderately confident this is a solved problem....

I've a 6 node cluster of dell r720xd's. each with 24 SSD OSDs (pve 7.3 upping to kernel 6.1 now..... ceph pacific.. (will go to quincy after living with 6.1 for a little bit)

What's the optimal way to evacuate vms/offline OSDs in a sane/lazy way?

I dislike the current behavior of 'pvenode migrateall' as it does't (at least as far as I can tell) have a way to say 'rebalance the cluster's VM distribution without factoring me into the mix'

In other words, I'm looking for a way to have a node express to the rest of its' clustermates

"I want to restart myself... can everyone else in the cluster take their share of my resources please? kthx!"

picking a node at random to shove everything to is kinda.... suboptimal.
I have several resource pools configured, each node is equally weighted... however resources don't seem to get relocated optimally...