[pgpool-general: 3838] Watchdog with Delegate IP Adress on Debian Jessie

[Marco Nietz]

Hi,

i try to get pgpool's (3.3.4-1) watchdog feature up and running on 
debian jessie (8.1) but experience problems as soon as i configure a 
virtual ip address in delegate_ip.

I have two servers.

pgpool01 (172.16.25.21)
pgpool02 (172.16.25.22)

With the following settings:

use_watchdog = on
wd_hostname = '172.16.25.21'
wd_port = 9000
wd_authkey = 'qwertz'
heartbeat_destination0 = '172.16.25.22'
other_pgpool_hostname0 = '172.16.25.22'
other_pgpool_port0 = 5432
other_wd_port0 = 9000

Watchdog starts on both nodes without any problems and even the state of 
the backend-servers is replicated. Watchdog service is bound, like 
expected to tcp-port 9000.

Logfile pgpool01:

pid 20050: watchdog might call network commands which using setuid bit.
pid 20050: wd_create_send_socket: connect() reports failure (Connection 
refused). You can safely ignore this while starting up.
pid 20050: send_packet_4_nodes: packet for 172.16.25.22:9000 is canceled
pid 20050: wd_escalation: escalating to master pgpool
pid 20050: wd_escalation: escalated to master pgpool successfully
pid 20050: wd_init: start watchdog
pid 20050: pgpool-II successfully started. version 3.3.4 (tokakiboshi)
pid 20050: find_primary_node: primary node id is 0
pid 20068: wd_create_hb_send_socket: set SO_REUSEPORT
pid 20068: wd_create_hb_recv_socket: set SO_REUSEPORT
pid 20070: wd_create_hb_send_socket: set SO_REUSEPORT
pid 20070: wd_create_hb_send_socket: set SO_REUSEPORT
pid 20067: wd_send_response: receive add request from 172.16.25.22:5432 
and accept it
pid 20071: watchdog: lifecheck started

Logfile pgpool02:

pid 7236: wd_chk_setuid all commands have setuid bit
pid 7236: watchdog might call network commands which using setuid bit.
pid 7236: wd_init: start watchdog
pid 7236: pgpool-II successfully started. version 3.3.4 (tokakiboshi)
pid 7236: find_primary_node: primary node id is 0
pid 7255: wd_create_hb_send_socket: set SO_REUSEPORT
pid 7255: wd_create_hb_recv_socket: set SO_REUSEPORT
pid 7256: wd_create_hb_send_socket: set SO_REUSEPORT
pid 7256: wd_create_hb_send_socket: set SO_REUSEPORT
pid 7258: watchdog: lifecheck started

Now i stop pgpool on both nodes and add a virtual ip-adress to the 
configuration file:

delegate_IP = '172.16.25.1'

Starting on pgpool01 results in:

pid 20169: wd_chk_setuid all commands have setuid bit
pid 20169: watchdog might call network commands which using setuid bit.
pid 20169: wd_create_send_socket: connect() reports failure (Connection 
refused). You can safely ignore this while starting up.
pid 20169: send_packet_4_nodes: packet for 172.16.25.22:9000 is canceled
pid 20169: wd_escalation: escalating to master pgpool

The ip-adress is started, but no watchdog is running on port 9000 and 
lifecheck isn't started.

Starting pgpool on the second node:

pid 7363: wd_chk_setuid all commands have setuid bit
pid 7363: watchdog might call network commands which using setuid bit.
pid 7363: wd_create_send_socket: connect() reports failure (Connection 
refused). You can safely ignore this while starting up.
pid 7363: send_packet_4_nodes: packet for 172.16.25.21:9000 is canceled
pid 7363: wd_init: delegate_IP 172.16.25.1 already exists
pid 7363: watchdog: wd_init failed
pid 7363: wd_main error
pid 7363: unlink(/var/run/postgresql/.s.PGSQL.9898) failed: No such file 
or directory
systemd[1]: pgpool2.service: main process exited, code=exited, 
status=1/FAILURE

It recognizes, that the virtual ip-adress is already started and exits.

Are there any know problems with this version on debian jessie ?


Best regards
Marco