[Pgpool-general] Error in pgpool

[Nimesh Satam]

All

We are facing problem with pgpool where the java application fails in
certain situation I/o problem.

When we see these problem we see in the db logs error statement like:

"ERROR:  prepared statement "S_3" does not exist".
"LOG:  unexpected EOF on client connection"

We get the following log output in the pgpool log files:

2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: Z
2009-03-17 12:24:56 DEBUG: pid 27834: pool_read_message_length: slot:
0 length: 5
2009-03-17 12:24:56 DEBUG: pid 27834: pool_read_message_length: slot:
1 length: 5
2009-03-17 12:24:56 DEBUG: pid 27834: ReadyForQuery: message length: 5
2009-03-17 12:24:56 DEBUG: pid 27834: ReadyForQuery: transaction state: I
2009-03-17 12:24:56 DEBUG: pid 27834: statement2: DEALLOCATE "S_3"
2009-03-17 12:24:56 DEBUG: pid 27834: SimpleQuery: set
master_slave_dml query: DEALLOCATE "S_3"
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: Query: BEGIN
2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: Z
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: transaction state: T
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: Query: BEGIN
2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: kind: Z
2009-03-17 12:24:56 DEBUG: pid 27834: do_command: transaction state: T
2009-03-17 12:24:56 DEBUG: pid 27834: waiting for backend 0 completing the query
2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: detect_error: kind: C
2009-03-17 12:24:56 DEBUG: pid 27834: waiting for backend 1 completing the query
2009-03-17 12:24:56 DEBUG: pid 27834: read_kind_from_backend: read
kind from 0 th backend C NUM_BACKENDS: 2
2009-03-17 12:24:56 DEBUG: pid 27834: read_kind_from_backend: read
kind from 1 th backend E NUM_BACKENDS: 2
2009-03-17 12:24:56 ERROR: pid 27834: read_kind_from_backend: 1 th
kind E does not match with master or majority connection k
ind C
2009-03-17 12:24:56 ERROR: pid 27834: kind mismatch among backends.
Possible last query was: "DEALLOCATE "S_3"" kind details
are: 0[C] 1[E]
2009-03-17 12:24:58 DEBUG: pid 27834: I am 27834 accept fd 5
2009-03-17 12:24:58 DEBUG: pid 27834: Protocol Major: 3 Minor: 0
database: rpt_production user: chakra
2009-03-17 12:24:58 DEBUG: pid 27834: new_connection: connecting 0 backend
2009-03-17 12:24:58 DEBUG: pid 27834: new_connection: connecting 1 backend
2009-03-17 12:24:58 DEBUG: pid 27834: pool_read_message_length: slot:
0 length: 8
2009-03-17 12:24:58 DEBUG: pid 27834: pool_read_message_length: slot:
1 length: 8
2009-03-17 12:24:58 DEBUG: pid 27834: pool_read_message_length2:
master slot: 0 length: 28
2009-03-17 12:24:58 DEBUG: pid 27834: pool_read_message_length2:
master slot: 1 length: 28
2009-03-17 12:24:58 DEBUG: pid 27834: 0 th backend: name:
client_encoding value: UNICODE
2009-03-17 12:24:58 DEBUG: pid 27834: 1 th backend: name:
client_encoding value: UNICODE

Why does pgpool try to DEALLOCATE a prepare statement on both the
database servers, when pgpool is set in Matser/slave mode. Can any
body high light how this issue can be solved. Will really appreciate
your inputs in this.

Regards,
Nimesh.