[ZOOKEEPER-4412] client blocked too long before session timeout

• Type: Bug
• Status: Open
• Priority: Critical
• Resolution: Unresolved
• Affects Version/s: 3.4.14
• Component/s: java client

Language

java

• Assignee: Unassigned
• Reporter: tian
• Created: 09/Nov/21 13:03
• Updated: 09/Nov/21 13:27

Description

set Negotiated session timeout: 40000

I have thousands of zk clients under a zk cluster, when switch failure, some of  clients did not send suspend event to state listner after 204988ms, but indeed session was expired in server as expected in server after 40000ms

// code placeholder
[10.10.10.2][my-app][2021-11-05 13:55:38.701][ WARN][????][16.195.152:2181)][o.a.z.ClientCnxn                        ] : [][][]Client session timed out, have not heard from server in 204988ms for sessionid 0x405005d85480c3d 
[10.10.10.2][my-app][2021-10-30 15:55:39.275][ WARN][????][16.195.152:218)][o.a.z.ClientCnxn                        ] : [][][]Unable to reconnect to ZooKeeper service, session 0x405005d85480c3d has expired 
[10.10.10.2][my-app][2021-10-30 15:55:39.275][ WARN][????][ain-EventThread][o.a.c.ConnectionState                   ] : [][][]Session expired event received

why after 204988ms this client found client session timeout ? socket read timeout = 2/3 * sessiont time out , it  blocked too long to discover network failure. what may cause this scene?