pgsql: report DISCONNECTED when it isn't

[kovur81]

Hi!
I`m trying to restore 2 node Pacemaker/Postgres cluster after h/w failure and facing an issue very similar to #104
My setup is

CentOS Linux release 7.9.2009 (Core)
NAME="CentOS Linux"
VERSION="7 (Core)"

Pacemaker 1.1.23-1.el7_9.1

postgresql12-server.x86_64 12.10-1PGDG.rhel7
systemctl is-enabled postgresql-12
disabled

pgsrv3 node Repl 172.18.0.1
pgsrv4 node Repl 172.18.0.2
Repl VIP is 172.18.0.3

crm_mon -Afr reports

Online: [ pgsrv3 pgsrv4 ]

Full list of resources:

Resource Group: master-group
masterIP (ocf::heartbeat:IPaddr2): Started pgsrv3
replIP (ocf::heartbeat:IPaddr2): Started pgsrv3
Master/Slave Set: pgsql-master-master [pgsql-master]
Masters: [ pgsrv3 ]
Slaves: [ pgsrv4 ]
SlaveIP (ocf::heartbeat:IPaddr2): Started pgsrv4

Node Attributes:

• Node pgsrv3:
  • master-pgsql-master : 1000
  • pgsql-master-data-status : LATEST
  • pgsql-master-master-baseline : 0000EEAFF60000A0
  • pgsql-master-status : PRI
  • pgsql-master-xlog-loc : 0000EEAFF60000A0
• Node pgsrv4:
  • master-pgsql-master : -INFINITY
  • pgsql-master-data-status : DISCONNECT
  • pgsql-master-status : HS:alone

but postgres reports

pgsrv3:

postgres=# select application_name,upper(state),upper(sync_state) from pg_stat_replication;
application_name | upper | upper
------------------+-----------+-------
walreceiver | STREAMING | ASYNC
(1 row)

When pgsrv3 fails pgsrv4 is not promoted to PRI accordingly

I'm bit of lost whats the problem with status of pgsrv4 under pacemaker, any help/advice will be very usefull.
Thank you!

[okamoto-sansan]

Try removing postgresql.auto.conf. can you get replication state?

pgsql looked for a node in node_list from application_name in ps_stat_replication status.

This application_name is not application_name in primary_conninfo, but cluster_name in postgresql.conf.

Checking postgresql.auto.conf, since application_name is not specified, the cluster name is adopted as the fallback application name.