I'm running v0.10.4. Very occasionally (three times out of thousands of pods), I've seen a case where, under high pod churn, a pod will start, but the csi driver will be rate limited by the client side rate limiter, and eventually fail to mount due to the 60 second timeout.
However, something happens and the main container starts anyway. The volume mount gets created, but without a cert, and it never gets issued at any later time either.
I don't think this is the expected behavior, and I suspect there is some race where an overlapping retry of the mount inadvertently marks the mount as successful.
I'm running v0.10.4. Very occasionally (three times out of thousands of pods), I've seen a case where, under high pod churn, a pod will start, but the csi driver will be rate limited by the client side rate limiter, and eventually fail to mount due to the 60 second timeout.
However, something happens and the main container starts anyway. The volume mount gets created, but without a cert, and it never gets issued at any later time either.
I don't think this is the expected behavior, and I suspect there is some race where an overlapping retry of the mount inadvertently marks the mount as successful.