Another controlplane is already in maintenance mod...
# harvester
r
Another controlplane is already in maintenance mode, cannot place current node in maintenance mode <- is this normal? i have five nodes, i set one to maintenance, shut it down, i can't set another one.
b
Yep pretty normal
It's trying to keep the Control Plane nodes in HA.
Let's say your nodes are: node1 node2 node3 node4 node5 Nodes 1-3 were the first in the cluster so they were elected/elevated to run the Control Plane. Only one of those can be "down" at a time. So you could put node4 and node5 into maintenance mode and the CP still has 3 nodes. You could put node3 out as well, but Harvester won't let you put node1 or node2 out until node3 is back.
It wouldn't matter if you have node4 or node5 back yet, because they're not running the kubeapi/etcd.
r
i see, is there a way around this ?
b
What are you trying to get around?
Like are you trying to shutdown nodes but have them come back online later? Are you trying to remove nodes? Shut down the cluster?
r
ya i was upgrading cpus, i found some cheap gold's to take out these silver's. so i tried to do two at a time, couldn't. i guess if i can't i can't
b
Not if they're both CP nodes
r
got it, so i think my guy just pulled server 2 while working on server 5 so i have now two boxes out of 5 up
not sure what happened, curious what issue this will cause
b
as long as one of the CP nodes is still up you should be ok
r
data is resilient up to how many failures/
b
You just need to keep an etcd leader
it can suffer 2/3 etcd nodes being down, but it'll start screaming.
r
ok let's see, wondering why it went down
harvester and cluster both notready on console, wonder what he did
oof
p2 i mean
maybe bad cpu
only thing different are other cpus
b
I'm not sure how Harvester deals with hardware/cpu changes
because part of the setup process is labeling the nodes with CPU specs
r
for the most part the hosts show the updated cpu and core count
just saw this server go down randomly though
b
You might try removing the nodes (one at a time) and kicking off a re-install
r
will have to see, one server is failing to post, other is failing , 3 are stable, maybe bad cpus, maybe the tech is bricked