StatefulSet pod can only be started with number 0. If the node with number 0 fails, the entire cluster will become unavailable.

[844700118]

StatefulSet pod can only be started with number 0. If the node with number 0 fails, the entire cluster will become unavailable.!

radondb-mysql version: v2.3.0 Kubernetes version (use kubectl version): 1.19 percona/percona-server:5.7.34 radondb/xenon:v2.3.0 radondb/mysql-operator:v2.3.0 radondb/mysql-sidecar:v2.3.0

[root@localhost ~]#  kubectl get sts -n storage-cluster  demo-mysql -o yaml
    apiVersion: apps/v1
    kind: StatefulSet
    metadata:
      name: ev-mysql
      namespace: storage-cluster
    spec:
      podManagementPolicy: OrderedReady
      replicas: 5

 [root@localhost ~]#  kubectl get pod -A -o wide|grep mysql
    storage-cluster ev-mysql-0  0/4  Pending 0 94m 100.113.172.1 master31 

    我的mysql集群有5个节点, 其中master31节点存在故障. 发现StatefulSet使用的是OrderedReady策略，
    如果编号0的Pod无法启动阻塞，那么编号1-4的Pod都无法启动，因为控制器会等待编号0的Pod变成Running和Ready后才会
    创建编号1的Pod. 请问 podManagementPolicy: OrderedReady 如果修改为 podManagementPolicy: Parallel.
    这样是否安全, 会不会导致选举出现问题. 有什么好的解决方案. 目的是集群高可用. 目前radondb集群是不满足的. 

    My mysql cluster has 5 nodes, among which the master31 node is faulty.
    I found that StatefulSet uses the OrderedReady strategy.
    If the Pod numbered 0 cannot start blocking, then the Pods numbered 1-4 cannot be started, 
    because the controller will wait for the Pod numbered 0 to become Running and Ready.
    Create Pod number 1. How to change podManagementPolicy: OrderedReady to podManagementPolicy: Parallel.
    Is this safe? Will it cause problems in the election? Is there any good solution? 
    The purpose is to make the cluster highly available. The current radondb cluster is not satisfied.

[github-actions[bot]]

Hi! thanks for your contribution! great first issue!

[acekingke]

You can try to label the node as taint node, so It will not schedule pod number 0 in it, It cannot use the Parallel policy, it must be very dangerous, causing the data inconsistency.