[pgpool-general: 4890] pgpool fail to restart after a pcp_stop_pgpool

[Daniel Huhardeaux]

I still have a problem to start a node. Setup is following: 2 servers, 
new start, TOOI-AS-0005 is master (node 0), TOOI-AS-0006 is standby 
(node 1).

I check replication so I created a table on pgpool delegate_IP. All is 
good, existing on both servers. I did a pcp_stop_pgpool on the master 
and created a table in my db, still using delegate_IP. All is OK, table 
is created which means that my standby switched to master.

I want now to attach the node 0 which was shutted down but can't. If I 
start a pg_pool_recovery -h delegate_IP 0 "I receive ERROR:  node 
recovery failed, node id: 0 is alive" If I try an pcp_attach_node -h 
delegate_IP 0, program hung, CTRL/C to exit.

I then restart pgpool with systemd (Ubuntu server 16.04) on node 0 and 
see the in my logs.

Aug  9 18:45:23 TOOI-AS-0005 pgpool[15510]: 2016-08-09 18:45:23: pid 
15510: FATAL:  failed to bind a socket: "/var/run/postgresql/.s.PGSQL.5432"
Aug  9 18:45:23 TOOI-AS-0005 pgpool[15510]: 2016-08-09 18:45:23: pid 
15510: DETAIL:  bind socket failed with error: "Address already in use"

After having removed manualy /var/run/postgresql/.s.PGSQL.5432, pgpool 
still doesn't start receving now

2016-08-09 19:06:49: pid 17899: FATAL:  failed to bind a socket: 
"/tmp/.s.PGSQL.9898"
2016-08-09 19:06:49: pid 17899: DETAIL:  bind socket failed with error: 
"Address already in use"

Removing manually/tmp/.s.PGSQL.9898 did pgpool starting properly again.

Could it be possible that pgpool remove those 2 files when we do a 
pcp_stop_pgpool ?

Best regards

-- 
TOOTAi Networks