https://rancher.com/ logo
Join Slack
Powered by
Having a weird one with upgrading from 31.7 to 33....
# rke2
s
Having a weird one with upgrading from 31.7 to 33.5 on my etcd/control nodes. When I start rke2-server, rke2-agent shutsdown, when I start rke2-agent, rke2-server shutsdown.
h
bit more context please? how did you upgrade? did you upgrade directly from 31.x to 33.x? what does journalctl show?
s
I did go straight to 33.x, bad me. Journalctl just shows the services being issued a kill command.
upgraded via rancher.
c
Yeah don't do that. Don't skip minors.
Also why are you trying to start the agent service on servers? you can't have agent and server services running on the same nodes. That has never worked or been supported.
The systemd services literally have a conflict hardcoded in them to prevent you from running both.
s
Well that would explain it.
c
server (control-plane/etcd) nodes run the server service. agent (worker) nodes run the agent service. why are you trying to start them both on the same node?
s
rke2-server and rke2-agent are both installed on the nodes, I assumed they should both be running.
The cluster was created via Rancher.
c
nope
if you created it via Rancher then let Rancher manage it, don’t go poking at things
s
I did let rancher manage it, It broke when I did a bad and upgraded by 2 versions instead of 1.
c
the install script drops systemd units for both node types, but only the one is ever enabled/started at once
s
Good to know, I appreciate it.
Copy code
INFO: Using resolv.conf: search cattle-system.svc.cluster.local svc.cluster.local cluster.local <http://xxxxxxxxx.xxx|xxxxxxxxx.xxx> nameserver 10.43.0.10 options ndots:5
ERROR: <https://xxxxxxxxxxxxxxxx.xxxxxxxxxxxxxx.xxx/ping> is not accessible (Could not resolve host: <http://xxxxxxxxxxxxxxx.xxxxxxxxx.xxx|xxxxxxxxxxxxxxx.xxxxxxxxx.xxx>)
Looks like my cluster register is not able to resolve the hostname of my rancher cluster. But every other container I check on that node can.
It looks like it failed during a helm update of coreDNS and so CoreDNS never installed.
lol..
It was systemd-resolved. I set the coreDNS config to point to upstream DNS and sure enough it started working.
3 Views
Open in Slack
PreviousNext
Powered by