View Issue Details

IDProjectCategoryView StatusLast Update
0000012Pgpool-IIBugpublic2012-05-10 05:08
ReporterpradavichAssigned To 
PrioritynormalSeverityminorReproducibilitysometimes
Status newResolutionopen 
PlatformLinuxOSCentOS release 5.6OS Version2.6.18-238.19.1.
Product Version 
Target VersionFixed in Version 
Summary0000012: Backend nodes can be considered failed during startup
DescriptionThe socket (like /tmp/.s.PGSQL.9898) pgpool uses can sometimes take a bit of time to close on linux. During a restart, this can happen:

May 9 17:57:32 localhost pgpool[16373]: pgpool-II successfully started. version 3.1.3 (hatsuiboshi)
May 9 17:57:32 localhost pgpool[16373]: bind(/tmp/.s.PGSQL.9898) failed. reason: Address already in use
May 9 17:57:32 localhost pgpool[16443]: connection received: host=apphost1 port=56894
May 9 17:57:32 localhost pgpool[16443]: connect_inet_domain_socket_by_port: exit request has been sent
May 9 17:57:32 localhost pgpool[16443]: connection to backend0 (5432) failed
May 9 17:57:32 localhost pgpool[16443]: new_connection: create_cp() failed
May 9 17:57:32 localhost pgpool[16443]: degenerate_backend_set: 1 fail over request from pid 16443
May 9 17:57:32 localhost pgpool[16463]: connection received: host=apphost2 port=60065
May 9 17:57:32 localhost pgpool[16461]: connection received: host=apphost3 port=56895
May 9 17:57:32 localhost pgpool[16463]: connect_inet_domain_socket_by_port: exit request has been sent
May 9 17:57:32 localhost pgpool[16461]: connect_inet_domain_socket_by_port: exit request has been sent
May 9 17:57:32 localhost pgpool[16461]: connection to backend0 (5432) failed
May 9 17:57:32 localhost pgpool[16461]: new_connection: create_cp() failed
May 9 17:57:32 localhost pgpool[16461]: degenerate_backend_set: 1 fail over request from pid 16461
May 9 17:57:32 localhost pgpool[16463]: connection to backend0 (5432) failed
May 9 17:57:32 localhost pgpool[16463]: new_connection: create_cp() failed
May 9 17:57:32 localhost pgpool[16463]: degenerate_backend_set: 1 fail over request from pid 16463

By this time, pgpool has exited because it cant bind the socket. It has also been unable to connect to the backend0 node, and marks it as failed in /tmp/pgpool_state.

When pgpool is started successfully after that, the backend0 node is still marked as failed.

This is with pgpool 3.1.3.

The backend node which is marked as failed can successfully be attached again.
Steps To ReproduceI can reproduce is pretty consistently by just running /etc/init.d/pgpool restart, using two backend nodes.

TagsNo tags attached.

Activities

There are no notes attached to this issue.

Issue History

Date Modified Username Field Change
2012-05-10 05:08 pradavich New Issue