[pgpool-general: 8490] pgpool shutdown due to [IN NETWORK TROUBLE]

[Steeve Boulanger]

Greetings,

We're facing an issue where an ansible playbook does a network reset on all
interfaces where we have pgpool running, and this seems to cause it to shut
down:

***
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network event
received
2022-11-08 18:22:39.737: watchdog pid 1323090: DETAIL:  deleted = YES Link
change event = NO
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  STATE MACHINE
INVOKED WITH EVENT = NETWORK IP IS REMOVED Current State = STANDBY
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network interface lo
having flags 65609
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network interface
ens192 having flags 4099
2022-11-08 18:22:39.737: watchdog pid 1323090: NOTICE:  network interface
"ens192" link is inactive
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network interface lo
having flags 65609
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network interface lo
having flags 65609
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  network interface
ens192 having flags 4099
2022-11-08 18:22:39.737: watchdog pid 1323090: NOTICE:  network interface
"ens192" link is inactive
2022-11-08 18:22:39.737: watchdog pid 1323090: WARNING:  network event has
occurred and all monitored interfaces are down
2022-11-08 18:22:39.737: watchdog pid 1323090: DETAIL:  changing the state
to in network trouble
2022-11-08 18:22:39.737: watchdog pid 1323090: LOG:  watchdog node state
changed from [STANDBY] to [IN NETWORK TROUBLE]
2022-11-08 18:22:39.737: watchdog pid 1323090: DEBUG:  STATE MACHINE
INVOKED WITH EVENT = STATE CHANGED Current State = IN NETWORK TROUBLE
2022-11-08 18:22:39.737: watchdog pid 1323090: FATAL:  system has lost the
network
2022-11-08 18:22:39.737: watchdog pid 1323090: LOG:  Watchdog is shutting
down
***
And since the playbook runs on all pgpool nodes, this is causing a DB
service outage.

I've looked at the documentation, but I haven't been able to find if there
was perhaps a timeout parameter, or some other way to avoid pgpool to shut
down during this short network reset. Any insights would be greatly
appreciated!

Thanks,
Steeve
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.pgpool.net/pipermail/pgpool-general/attachments/20221108/268d0d6b/attachment.htm>