https://rancher.com/ logo
Join Slack
Powered by
This message was deleted.
# rke2
a
This message was deleted.
n
i used the same bootstrap ip (https://10.10.5.10:9345) as i used for all agents
c
look for other log messages about adding a new member to the etcd cluster. make sure you’ve opened the etcd ports between nodes. if you only had agents before its likely you never had to open any ports for this.
port requirements are covered in the docs
n
the nodes should be able to connect to any port but i could try telnet - which port would that be?
c
n
so from the new server, i should connect to server1 on port 2380
refused - ok i guess i have some firewall issues. Thank you so much!
c
do NOT have 2 servers though
2 servers is worse than 1
have 1 or 3
n
so quorum is the issue you assume?
current i actually just migrate the cluster. So basically the old 3 workers are replaced by new workers, same with the server- But i guess those servers will fight for quorum? Can i configure that 1 vote is enouhg (for now)
c
if you have two etcd nodes and either one is down, the other one will not work either. due to quorum.
n
so there is no good way to run with 2, even by setting the vote min to 1? (so 50% is ok)
nevertheless, seem like my server1 node might be a little "confused" about the node-ip. Seems like the api server is running oin my first ip (10.10.5.10) but the etcd is running on the second ip 10.10.5.50 - i assume all i need to do i settin node-ip in the config.yaml and restart the server?
somehow i broke the cluster on server1. I had 10.10.5.10 in the sans and added 10.10.5.50, then run rke2 certificates rotate and restarted the server
level=info msg="Failed to test temporary data store connection: etcd datastore is not started"
i might have forgotted to run 
rke2 certifcate rotata-ca
ofter running 
rke2 certificate rotate
?
restored the old tls certs and got the server1 cluster back up - so back at square 1
i was now be able add the new server. Now checking how to configure the quorum
c
I really wouldn’t do that
n
@creamy-pencil-82913 not ignoring your advice at all - what i'am doing right now is more or less trying to migrate from one section of the cluster (old) to the new - that's why i have 2 planes and 6 workers. My plan is to get them all running and then cordon + drain the old agents, so that payload is moved to the new agents. Do that 3 times so all old agents are drained - then remove them
After that, i plan to switch the control plane from old to new, but for this i need to have them both registerd and clone the cluster state
Then i would drain/remove the old control plane and run with 1/3 again. I understood that if i want HA on the control planes, i will need to have 3 of those
Do you suggest, that even during this "moving" effort, i should add a third server?
57 Views
Open in Slack
PreviousNext
Powered by