[Pgpool-general] Replication conncetions failing; help with debugging?

[Josh Berkus]

So, replication mode with pgpool is consistently blowing up on me. I'd
like some help with debugging why.  Here's what's happening.

On the client:
Connection for session 0x4bf5867000003b70 failed: server closed the
connection unexpectedly
        This probably means the server terminated abnormally
        before or while processing the request.

On the pgpool server:

2010-08-27 21:00:16 DEBUG: pid 30231: pool_process_query: kind from
backend: C
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 0 th backend Z NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 1 th backend Z NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: pool_process_query: kind from
backend: Z
2010-08-27 21:00:16 DEBUG: pid 30231: pool_read_message_length: slot: 0
length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: pool_read_message_length: slot: 1
length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: ReadyForQuery: message length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: ReadyForQuery: transaction state: I
2010-08-27 21:00:16 LOG:   pid 30231: ProcessFrontendResponse: failed to
read kind from frontend. frontend abnormally exited
2010-08-27 21:00:16 DEBUG: pid 30231: statement2:  DISCARD ALL
2010-08-27 21:00:16 DEBUG: pid 30231: wait_for_query_response: waiting
for backend 0 completing the query
2010-08-27 21:00:16 DEBUG: pid 30231: detect_error: kind: S
2010-08-27 21:00:16 DEBUG: pid 30231: detect_error: kind: S
2010-08-27 21:00:16 DEBUG: pid 30231: detect_error: kind: S
2010-08-27 21:00:16 DEBUG: pid 30231: detect_error: kind: S
2010-08-27 21:00:16 DEBUG: pid 30231: wait_for_query_response: waiting
for backend 1 completing the query
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: parameter
name: is_superuser value: off
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: parameter
name: session_authorization value: at
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 0 th backend C NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: parameter
name: is_superuser value: off
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: parameter
name: session_authorization value: at
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 1 th backend C NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: pool_process_query: kind from
backend: C
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 0 th backend Z NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: read_kind_from_backend: read kind
from 1 th backend Z NUM_BACKENDS: 2
2010-08-27 21:00:16 DEBUG: pid 30231: pool_process_query: kind from
backend: Z
2010-08-27 21:00:16 DEBUG: pid 30231: pool_read_message_length: slot: 0
length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: pool_read_message_length: slot: 1
length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: ReadyForQuery: message length: 5
2010-08-27 21:00:16 DEBUG: pid 30231: ReadyForQuery: transaction state: I
2010-08-27 21:00:16 DEBUG: pid 30231: pool_connection_pool_timer: set
close time 1282957216

The database server logs show no errors.  Also, note that this
particular set of servers worked perfectly in master-slave mode with
slony doing the replication.

-- 
                                  -- Josh Berkus
                                     PostgreSQL Experts Inc.
                                     http://www.pgexperts.com