[Pgpool-general] Able to do online recovery for failed nodes but not to attach new node

[Lao Wai]

I think there is a problem in pgpool reloading the config file. If I create the config file to have a valid backend0 but an invalid hostname for backend1:

backend_weight0 = 1
backend_port0 = 5432
backend_hostname0 = domU-12-31-39-00-58-57.compute-1.internal
backend_data_directory0 = '/vol/postgresdata'
backend_weight1 = 1
backend_port1 = 5432
backend_hostname1 = foo
backend_data_directory1 = '/vol/postgresdata'

And I launch pgpool. It immediately detects that backend1 is down and fails over. I see period health checks:

2009-04-22 07:17:42 DEBUG: pid 1437: health_check: 0 th DB node status: 1
2009-04-22 07:17:42 DEBUG: pid 1437: health_check: 1 th DB node status: 3

Fine. Now I update pgpool.conf:

backend_hostname1 = bar

Now I issue pgpool reload:

2009-04-22 07:21:17 DEBUG: pid 1510: key: backend_hostname1
2009-04-22 07:21:17 DEBUG: pid 1510: value: bar kind: 1

Now I run pcp_node_info on node 1:

foo 5432 3 1073741823.500000

It's still the old value, even though the debug output shows loading the previous value. I'm pretty sure this is the root of the problem, issuing reload is not fully committing the new update.

Anyone have any clues/suggestions?

_________________________________________________________________
Windows Live™ Hotmail®:…more than just e-mail.
http://windowslive.com/online/hotmail?ocid=TXT_TAGLM_WL_HM_more_042009
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://pgfoundry.org/pipermail/pgpool-general/attachments/20090422/88d8e656/attachment.html