[Pgpool-general] Regarding the ProcessFrontendResponse error

[Jaume Sabater]

Hello everyone!

I've been testing pgpool-II RC1 with the createindex patch this
morning, using PostgreSQL 8.3.6 on AMD64 architecture.

The last test I've doing is to grab a programmer and tell her to make
a reservation in the system (online sales of hotel bookings). During
the process, four SELECT queries were executed against the PostgreSQL
cluster (all transactions are stored in a different server, being
PostgreSQL used only for reading information).

After the reservation was complete, I found this:

a) PostgreSQL 1 log:

2009-02-19 12:07:51 CET LOG:  statement:  DISCARD ALL

b) PostgreSQL 2 log:

2009-02-19 12:07:51 CET LOG:  statement:  DISCARD ALL

c) pgpool-II log (syslog):

Feb 19 12:07:51 pgsql1 pgpool: ERROR: pid 6420: pool_read: read failed
(Connection reset by peer)
Feb 19 12:07:51 pgsql1 pgpool: LOG:   pid 6420:
ProcessFrontendResponse: failed to read kind from frontend. frontend
abnormally exited

After each SELECT query the front-end executed, I found these on the
PostgreSQL log (on both nodes):

2009-02-19 12:07:26 CET LOG:  statement: BEGIN
2009-02-19 12:07:26 CET LOG:  statement: deallocate "npgsqlplan1";
2009-02-19 12:07:26 CET LOG:  statement: COMMIT
2009-02-19 12:07:26 CET LOG:  statement: BEGIN
2009-02-19 12:07:26 CET LOG:  statement: unlisten *
2009-02-19 12:07:26 CET LOG:  statement: COMMIT

But these did not produce any error. Now I've been talking to two
programmers and they have told me they are using the npgsql driver
version 2.0.2 from http://pgfoundry.org/projects/npgsql/. We all
believe the stuff posted above must be the driver's fault, but I
wanted you guys to give me your opinion. I don't like having error
messages on my log, even if they "can be ignored".

Any ideas?

-- 
Jaume Sabater
http://linuxsilo.net/

"Ubi sapientas ibi libertas"