Closed jellonek closed 5 years ago
Also this change should simplify debugging in situations like described in #825
As discussed internally - CrashLoopBackoff should automatically throttle down creation mounts explosion and --previous
parameter for kubectl logs
should resolve looking into old logs, without a need to add ugly sleep infinite
.
At the moment if there is an issue with prepare node, which is an init container - whole pod will be restarted and in effect reaching the logs with description of failure is quite hard.
This change should simplify the way of gaining information about cause of failure while also will allow to continue starting procedure if a system administrator would fix the issue manually.
This change is