[pgpool-hackers: 3503] weird problems: our stand for coordinator request is rejected by node

Sebastian sebastian at otherstuff.nl
Thu Jan 30 00:41:52 JST 2020


Hi,

Due to an templating error I ran into stupid but very hard to debug
error, which might want catching with a warning.

The error is caused by heaving also itself in the other_pgpool
configuration (which happened due to an ansible error :).

The result however when you start the first pgpool instance, it doesn't
want to be the master, with an output as such:

2020-01-29 14:37:11: pid 57802: LOG:  pool_discard_oid_maps: discarded
memqcache oid maps
2020-01-29 14:37:11: pid 57802: LOG:  waiting for watchdog to initialize
2020-01-29 14:37:11: pid 57805: LOG:  setting the local watchdog node
name to "<hostname1>:5432 Linux <hostname1>"
2020-01-29 14:37:11: pid 57805: LOG:  watchdog cluster is configured
with 2 remote nodes
2020-01-29 14:37:11: pid 57805: LOG:  watchdog remote node:0 on
<hostname1>:9000
2020-01-29 14:37:11: pid 57805: LOG:  watchdog remote node:1 on
<hostname2>:9000
2020-01-29 14:37:11: pid 57805: LOG:  interface monitoring is disabled
in watchdog
2020-01-29 14:37:11: pid 57805: LOG:  watchdog node state changed from
[DEAD] to [LOADING]
2020-01-29 14:37:11: pid 57805: LOG:  new watchdog node connection is
received from "192.168.151.12:25768"
2020-01-29 14:37:11: pid 57805: LOG:  new outbound connection to
<hostname1>:9000
2020-01-29 14:37:11: pid 57805: LOG:  new node joined the cluster
hostname:"<hostname1>" port:9000 pgpool_port:5432
2020-01-29 14:37:11: pid 57805: DETAIL:  Pgpool-II version:"4.1.0"
watchdog messaging version: 1.1
2020-01-29 14:37:16: pid 57805: LOG:  watchdog node state changed from
[LOADING] to [JOINING]
2020-01-29 14:37:16: pid 57805: LOG:  watchdog node state changed from
[JOINING] to [INITIALIZING]
2020-01-29 14:37:17: pid 57805: LOG:  watchdog node state changed from
[INITIALIZING] to [STANDING FOR MASTER]
2020-01-29 14:37:17: pid 57805: LOG:  our stand for coordinator request
is rejected by node "<hostname1>:5432 Linux <hostname1>"
2020-01-29 14:37:17: pid 57805: LOG:  watchdog node state changed from
[STANDING FOR MASTER] to [PARTICIPATING IN ELECTION]
2020-01-29 14:37:22: pid 57805: LOG:  watchdog node state changed from
[PARTICIPATING IN ELECTION] to [JOINING]
2020-01-29 14:37:22: pid 57805: LOG:  watchdog node state changed from
[JOINING] to [INITIALIZING]
and then looping through the last bits.

The confusing bit here is that it not clear why it's being rejected. But
perhaps a check on whether a remote pgpool is the same as the running
pgpool might solve a lot of headaches.

This was all tested on pgpool2  version 4.1.0-1.pgdg90+1 running on
Debian stretch/9. Using a 2 instance pgpool configuration (but the
problem exists with just one :).

Hope this is of help to anyone, thanks!

-- 
Sebastian Stellingwerff
06-49136664


More information about the pgpool-hackers mailing list