View Issue Details

IDProjectCategoryView StatusLast Update
0000409Pgpool-IIBugpublic2018-07-23 10:37
Reporternagata Assigned ToMuhammad Usama  
PrioritynormalSeverityminorReproducibilitysometimes
Status closedResolutionfixed 
Product Version3.7.4 
Summary0000409: worker process is not restarted after failover on standby pgpool with in watchdog cluster
DescriptionWhen a backend goes down, failover is performed by the master pgpool, and the backend status is sent to the standby node. Then, the standby node's backend status is altered by sync_backend_from_watchdog(). In this function, child processes are restarted if needed, but worker process is not restarted. It cause the following errors continuing repeatedly.

ERROR: failed to make persistent db connection
ERROR: Failed to check replication time lag

To fix this, worker process have to be restarted in the end of sync_backend_from_watchdog(), as the attached patch.
Steps To ReproduceConfigure streaming replication + watchdog cluster with two pgpool-II and two backend.
health_check_period and sr_check_period must be >0.

After starting the cluster, shout down the standby DB by pg_ctl stop -m fast

The standby pgpool-II will report "Failed to check replication time lag" errors repeatedly.
TagsNo tags attached.

Activities

nagata

2018-07-02 20:39

developer  

Muhammad Usama

2018-07-03 21:43

developer   ~0002084

Hi
Thanks for the patch, I have committed to all related branches.
https://git.postgresql.org/gitweb/?p=pgpool2.git;a=commitdiff;h=bbdf0d92483caff9954a91bf5156710fc9f5ceff

Issue History

Date Modified Username Field Change
2018-07-02 20:37 nagata New Issue
2018-07-02 20:37 nagata File Added: respawn_worker_process.patch
2018-07-02 20:38 nagata File Deleted: respawn_worker_process.patch
2018-07-02 20:39 nagata File Added: respawn_worker_process.patch
2018-07-03 21:43 Muhammad Usama Assigned To => Muhammad Usama
2018-07-03 21:43 Muhammad Usama Status new => resolved
2018-07-03 21:43 Muhammad Usama Resolution open => fixed
2018-07-03 21:43 Muhammad Usama Note Added: 0002084
2018-07-23 10:37 administrator Status resolved => closed