[pgpool-general: 6534] Not quite getting what "replication delay" is representing ... ?

[Martin Goodson]

Hello.

I am attempting to configure an existing PostgreSQL 9.5.6 database to 
use pgpool 4.0.4, and I'm hitting something a bit curious.

The database is running in pretty standard streaming replication mode, 
with the primary on one server and the standby on a second server. As 
far as I can see replication is working well, and doesn't appear to be 
particularly slow. The pgpool configuration file is set with 
master_slave_mode = on and master_slave_sub_mode = "streaming" with 
replication_mode = off, load_balance_mode = off.

If I test replication on the primary and standby, by for example running 
an update on the primary and looking at the result on the standby, I can 
see that the streaming replication is pretty quick - guaranteed 
subsecond every time I look at it.

Yet if I do a show pool_nodes on the pgpool server I see extensive 
'replication delay'. And, even when the database is not doing anything 
that figure remains high.

If I look in the pgpool log file I see a number of messages like this:

LOG: Replication of node:1 is behind 1458585952 bytes from the primary 
server (node:0)

1,458,585,952 bytes? Nooo, not on the database. If I look in 
pg_stat_replication all four locations (sent, write, flush, replay) are 
the same. As far as the database is concerned there's no appreciable delay.

Am I missing something blindingly obvious here? Have I forgotten to set 
an option somewhere and this delay ... isn't? Is this something I can 
ignore with replication mode = off? If so, is there a way to disable the 
display of these figures in the log?

Many thanks, and apologies if these are stupid questions :)

Regards,

Martin.

-- 
Martin Goodson

"Have you thought up some clever plan, Doctor?"
"Yes, Jamie, I believe I have."
"What're you going to do?"
"Bung a rock at it."