[BUG] The number of unavailable instances may exceed the maxUnavailable limit during upgrades

[YTGhost]

Describe the bug When performing an upgrade, if the user upgrades to a faulty image causing the Container to crash shortly thereafter, during this brief period (before the Container crashes), since the Pod has already been upgraded, the next Pod upgrade might be permitted (with currentUnavailable being 0 and the previous Pod being checked as healthy). This can lead to the final number of unavailable instances exceeding the maxUnavailable limit.

Expected behavior In addition to the isHealthy check, it may also be necessary to rely on MinReadySeconds for the isRunningAndAvailable check. At the same time, we also need to ensure that the call to isRunningAndAvailable happens after the kubelet has completed updating the status.