[Pgcluster-general] stand alone node with read_write mode problem

[Marcin Gabrowski]

Hello,

I've configured pgcluster 1.9rc5 as follows:
1st server:  pgreplicate, clusterdb (pg 8.3)
2nd server: pgreplicate, clusterdb (pg 8.3)
3rd server: pglb,  java aplication
4th server: pglb,  java aplication

I've set read_write mode on each stand alone node when others fails.
The problem is start transaction in above state as we can see in logs  
of pgreplicate on alive machine:

Fri Mar 28 11:33:33 2008  dbName=template1
Fri Mar 28 11:33:33 2008  userName=pgcluster
Fri Mar 28 11:33:33 2008  recieve sec=0
Fri Mar 28 11:33:33 2008  recieve usec=0
Fri Mar 28 11:33:33 2008  query_size=63
Fri Mar 28 11:33:33 2008  request_id=0
Fri Mar 28 11:33:33 2008  replicate_id=0
Fri Mar 28 11:33:33 2008  recovery_status=0
Fri Mar 28 11:33:33 2008  query=SELECT  
PGR_SYSTEM_COMMAND_FUNCTION(1,'192.168.200.22',8001,8101)
Fri Mar 28 11:33:33 2008  PGRis_same_host():target host
Fri Mar 28 11:33:33 2008  PGRcreateConn():Retry. h_errno is 1,reason  
is 'FATAL:  the database system is in recovery mode'
Fri Mar 28 11:33:33 2008  PGRcreateConn():Retry. h_errno is 1,reason  
is 'FATAL:  the database system is in recovery mode'
Fri Mar 28 11:33:33 2008  PGRcreateConn():Retry. h_errno is 1,reason  
is 'FATAL:  the database system is in recovery mode'
Fri Mar 28 11:33:33 2008  PGRcreateConn():Retry. h_errno is 1,reason  
is 'FATAL:  the database system is in recovery mode'
Fri Mar 28 11:33:33 2008  setTransactionTbl(): 5830 at 192.168.200.21 is  
not ready
Fri Mar 28 11:33:33 2008   
replicate_packet_send_internal():setTransactionTbl failed
Fri Mar 28 11:33:33 2008  PGRis_same_host():target host
Fri Mar 28 11:33:33 2008  start thread_send_cluster()
Fri Mar 28 11:33:33 2008  send_replicate_packet_to_server():PQexec  
send :SELECT PGR_SYSTEM_COMMAND_FUNCTION(1,'192.168.200.22',8001,8101)
Fri Mar 28 11:33:33 2008  send_replicate_packet_to_server():PQexec  
returns :SYSTEM_COMMAND
Fri Mar 28 11:33:33 2008  thread_send_cluster():return value from  
send_replicate_packet_to_server() is 0
Fri Mar 28 11:33:33 2008  thread_send_cluster():pthread_exit[0]

so,
postgres=# BEGIN ;

(many sec)

Cancel request sent
Cancel request sent

I supose that degrade mode will allow me write to last node of  
pgcluster db.

-- 
gg: 2600 ripe: mg3051 jid: gaber/gentoo.pl http://gabrowski.pl/