redis sentinel 错误处理

环境

redis-ha主从部署到了k8s上

主从报错日志

主库日志

1:M 29 Nov 2023 07:00:19.277 # Diskless rdb transfer, done reading from pipe, 1 replicas still up.
1:M 29 Nov 2023 07:00:20.058 * Background RDB transfer terminated with success
1:M 29 Nov 2023 07:00:20.058 * Streamed RDB transfer with replica 10.97.141.243:6379 succeeded (socket). Waiting for REPLCONF ACK from slave to enable streaming
1:M 29 Nov 2023 07:00:20.058 * Synchronization with replica 10.97.141.243:6379 succeeded
1:M 29 Nov 2023 07:00:20.058 * Starting BGSAVE for SYNC with target: replicas sockets
1:M 29 Nov 2023 07:00:20.691 * Background RDB transfer started by pid 897
1:M 29 Nov 2023 07:02:50.161 # Connection with replica 10.97.141.243:6379 lost.

从报错可知，是开启了无盘复制参数 repl-diskless-sync

从库日志

1:S 29 Nov 2023 06:54:45.961 * MASTER <-> REPLICA sync: receiving streamed RDB from master with EOF to disk
1:S 29 Nov 2023 07:00:18.665 * MASTER <-> REPLICA sync: Flushing old data
1:signal-handler (1701241365) Received SIGTERM scheduling shutdown...

经过一系列的排查未发现 redis 及内存方面的原因导致这种情况出现，最后describe pod 的时候发现存活性检测失败的问题，进一步排查发现从库在 Flushing old data 的时候会阻塞那个存活性检测脚本，导致失败，然后 kill 掉 pod，固日志中会有 Received SIGTERM scheduling shutdown 这个信号。

解决方法

将 livenessProbe 中 timeoutSeconds 调大即可。

  livenessProbe:
    initialDelaySeconds: 30
    periodSeconds: 15
    timeoutSeconds: 900
    successThreshold: 1
    failureThreshold: 5