zk/keeper restarts cause zk watch to be unbalanced
Company or project name
Company:Tencent
Question
Discussion: Restart zk/keeper will lead to the imbalance of zk watch, which will cause high pressure on the single node, resulting in long writing and merge time. Could you increase SYSTEM to force the reconnection of zk/keeper.
Which version did you use?
Which version did you use?
21.8 or 22.8 or 23.8 , three versions have this problem
Version 23.8 should support fallback_session_lifetime parameter to rebalance the ZK connections.
This is improved even more in 24.7
See the fallback_session_lifetime min and max values in the docs.
Aside from that, you can restart the clickhouse cluster after the keeper/zk restart.