mhagander / bgw_replstatus Goto Github PK

I noticed that @snopoke added some commits to check for replication lag on a standby server. That could be helpful for a haproxy configuration where you just need a somewhat "current" read-only instance.

Do you think integrating these commits makes sense?

hi everyone.
I try to implement PostgreSQL high availability using bgw_replstatus and haproxy.
according to part of your readme, it said :

automatically switches over to the backup node if the master goes down and the backup is promoted.

but I have an issue when the master instance is down, the slave does not take over/ promoted to master node. here is my environment :

1. dockerized PostgreSQL 9.68
2. installed bgw_replstatus version 1.02 inside the container fetch from
3. haproxy 1.7.11-1

here is my postgresql.conf configuration on master node that I change :

wal_level = hot_standby
archive_mode = on
archive_command = 'test ! -f mnt/server/archivedir/%f && cp %p mnt/server/archivedir/%f'
max_wal_senders = 3
wal_keep_segments = 1000
shared_preload_libraries = 'bgw_replstatus'

here is my postgresql.conf on slave node after pg_basebackup which I change:

hot_standby = on

and here is my recovery.conf on the slave node

standby_mode = on
primary_conninfo = 'host=[host-ip-source] port=[source-port] user=repuser password=[some-password]'

and here is my haproxy configuration

frontend pgcluster
	bind *:5433
	mode tcp
	default_backend pgcluster

backend pgcluster
	mode tcp
	balance roundrobin
	option tcp-check
	tcp-check expect string MASTER
	server tm 172.31.32.129:5444 check port 5446
        server x2 172.31.18.158:5554 check port 5400 backup

here is the return when I test it using nc [ip-address] [port-bgw_replstatus] command

when all node is up, the PostgreSQL running well, and I can access it using this command :
psql -h some.domain.to.pgcluster -p 5433 -U postgres
 

but when the master is down, the slave is not promoted to master/not takeover here is the return :

is there a miss configuration or is there workaround which I can use to implement this high availability for haproxy?