[Pgpool-general] What will happen, when secondary node is down
Qian Xu
qian.xu at funkwerk-itk.com
Mon Feb 16 17:34:41 UTC 2009
Hi All,
could you tell me, what will happen, when the secondary node is down?
I have deployed pgpool on two computers with postgres (v8.1).
Currently the secondary node is down. (master=up, secondary=down)
I just wondered, pgpool works without big issues.
---------- important configuration -------------
replication_mode = true
replication_timeout = 60000
load_balance_mode = true
master_slave_mode = false
connection_cache = false
health_check_period = 0
-------------------------------------------------
But sometimes, database performs very slow by executing "update
table_name ....".
It will take more than 60 seconds. (no idea why so slow)
And after 60 seconds, the pool will be resynchronized, because of the
timeout.
Here is a dumped log:
---------- dump ------------------------------------------------
2009-02-11 14:36:51 ERROR: pid 17108: pool_check_fd: data is not ready
tp->tv_sec 60 tp->tp_usec 60000000 for peer 20.1.84.132:5433
2009-02-11 14:36:51 ERROR: pid 17108: pool_process_query: secondary data
is not ready at synchronous point. abort this session
2009-02-11 14:36:51 ERROR: pid 17108: line 578: pool_process_query: kind
does not match between backends master(C) secondary( )
2009-02-11 14:36:51 DEBUG: pid 17108: connected to
BBZB-RSS-SR-MASTER:5433 (local:42368)
2009-02-11 14:36:51 DEBUG: pid 17108: connected to
BBZB-RSS-SR-SLAVE:5433 (local:50116)
2009-02-11 14:36:51 LOG: pid 17108: pool_process_query: (line 625):
backend (-1) will be set to down due to the resync problem
2009-02-11 14:36:51 LOG: pid 17108: notice_backend_error: master: -1
fail over request from pid 17108 (conn: 0)
2009-02-11 14:36:51 LOG: pid 17108: do_child: no exit at this point,
but notification of error to parent is done for 20.1.84.128:40025
2009-02-11 14:36:51 DEBUG: pid 23725: basic failover_handler called by
signal 10 with suppress option 0 (0 == no suppress)
2009-02-11 14:36:51 LOG: pid 23725: starting degeneration. shutdown
master host BBZB-RSS-SR-01(5433)
---------- dump -------------
According to the dump file. The secondary node sent ( ) back, instead of
(C).
I think this is incorrect.
Can someone explain, how will pgpool work when the secondary node is
down.
Any hint to solve my problem will be very appreciated.
Best regards
--Qian Xu
More information about the Pgpool-general
mailing list