Hi Team, I've encountered an interesting behavior ...
# rke2
m
Hi Team, I've encountered an interesting behavior in my RKE2 cluster during updates to the FQDN configurations on my server nodes and would appreciate insights or explanations. Scenario: • My cluster consists of several server nodes, notably
server0
and
server1, server2
. • Recently, I updated the FQDN configuration for these servers.
server0
does not have a server entry in its RKE2 configuration, while
server1
does. • After applying the new FQDN config and restarting
server0
for the first time, I observed that other nodes in the cluster temporarily transitioned to a
NotReady
state before returning to
Ready
. • Interestingly, performing the same operation first on
server1
(which has a server entry in its config) does not cause the other nodes to enter a
NotReady
state. Questions: 1. Why does restarting
server0
(without a server entry) cause other nodes to temporarily become
NotReady
, whereas restarting
server1
(with a server entry) does not have the same effect? 2. Are there specific considerations or best practices when updating FQDN configurations in RKE2 clusters, especially concerning the server entries in the config files? 3. Could this behavior be related to how the RKE2 cluster manages internal DNS resolutions or certificate trust chains upon FQDN changes? Any guidance on understanding and ideally managing this behavior would be greatly appreciated. I'm aiming to ensure minimal impact on the cluster's availability and state consistency during configuration updates.