View Issue Details

Jump to NotesJump to History
IDProjectCategoryView StatusDate SubmittedLast Update
0000591Pgpool-IIBugpublic2020-03-04 15:182020-03-06 10:24
Reporternervzhang Assigned Tot-ishii  
PrioritynormalSeveritymajorReproducibilityalways
Status closedResolutionopen 
PlatformLinuxOSCentOSOS Version7.7.1908
Product Version4.1.1 
Summary0000591: pgpoo stuck under a large amount of concurrency
Descriptionwhen I do stress tests,if the concurrency is higher than the num_init_children,pgpool will stuck until I shutdown the test.

After the test close,pgpool show me following information:
2020-03-04 11:46:04: pid 2658: ERROR: unable to read data from frontend
2020-03-04 11:46:04: pid 2658: DETAIL: EOF encountered with frontend

Steps To Reproducepostgresql master node：192.168.100.1
postgresql slave1 node：192.168.100.11
postgresql slave2 node：192.168.100.12
pgpool：192.168.100.20
tester：192.168.100.60

All nodes are virtual machine,and connected by internal switch.

1.start postgresql nodes and pgpool node as normal
2.run "sudo -u postgres /usr/pgsql-12/bin/pgbench -h 192.168.100.20 -t 20 -c 250 -j 16 -r postgres -p 9999 -f select.sql" from tester node
TagsNo tags attached.

Activities

nervzhang

2020-03-04 15:18

reporter  

pgpool.conf (43,774 bytes)
postgresql.conf (26,697 bytes)
select.sql (59 bytes)

t-ishii

2020-03-04 16:03

developer   ~0003239

It's normal. See the FAQ:
https://pgpool.net/mediawiki/index.php/FAQ#When_I_run_pgbench_to_test_pgpool-II.2C_pgbench_hangs._If_I_directly_run_pgbench_against_PostgreSQL.2C_it_works_fine._Why.3F

nervzhang

2020-03-04 16:22

reporter   ~0003240

My concurrency value is 250,num_init_children value is 180,postgresql max_connections value is 300.In my opnion,first batch of 180 connections will execute correctly,and then when first batch done,second batch of 70 connections will execute from the queue.Could you tell me where is wrong,thanks.

t-ishii

2020-03-04 16:51

developer   ~0003241

No, that's not how pgbench works. Until all of 250 connections is established, no transaction will be issued by pgbench. So "first batch" will never be executed.

nervzhang

2020-03-04 16:53

reporter   ~0003242

Thanks a lot.

t-ishii

2020-03-05 17:28

developer   ~0003245

May I close this issue?

nervzhang

2020-03-05 17:30

reporter   ~0003246

Yes thanks

Issue History

Date Modified Username Field Change
2020-03-04 15:18 nervzhang New Issue
2020-03-04 15:18 nervzhang File Added: pgpool.conf
2020-03-04 15:18 nervzhang File Added: postgresql.conf
2020-03-04 15:18 nervzhang File Added: select.sql
2020-03-04 16:03 t-ishii Assigned To => t-ishii
2020-03-04 16:03 t-ishii Status new => feedback
2020-03-04 16:03 t-ishii Description Updated
2020-03-04 16:03 t-ishii Steps to Reproduce Updated
2020-03-04 16:03 t-ishii Note Added: 0003239
2020-03-04 16:22 nervzhang Note Added: 0003240
2020-03-04 16:22 nervzhang Status feedback => assigned
2020-03-04 16:51 t-ishii Note Added: 0003241
2020-03-04 16:53 nervzhang Note Added: 0003242
2020-03-05 17:28 t-ishii Note Added: 0003245
2020-03-05 17:28 t-ishii Status assigned => feedback
2020-03-05 17:30 nervzhang Note Added: 0003246
2020-03-05 17:30 nervzhang Status feedback => assigned
2020-03-06 10:24 t-ishii Status assigned => closed