Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

Performance issues in pstress

Description

While working on pstress, several performance issues are noticed.

1. threads are under utilised, show processlist reveals not all threads are executing queries

2. threads simply wait and prints message in the logs "
Waiting for all threds to finish initial load
Why not utilise waiting threads to create pending metadata.

3. On QA machines, a load of 100 tables with 5000 records is taking more than 15 mins. sysbench on the same machine does the same task in much lesser time.

To repeat, use driver script script to execute load
DYNAMIC_QUERY_PARAMETER="--tables 50 --records 5000 --log-all-queries --no-select --exact-initial-records --no-temp-tables"
PSTRESS_RUN_TIMEOUT=300

Environment

None

Smart Checklist

Activity

mohit.joshi May 8, 2024 at 5:56 AM

commit c99ef4cf700dad5bffff56c959422ffeb1c8d061
Author: Rahul Malik rahul.malik@percona.com
Date: Thu Oct 20 15:09:29 2022 +0530

Rahul Malik September 15, 2022 at 10:10 AM

for performace we need to fix 2 things.

1) create a secondary index after inserting the initial records
2) do bulk insert instead of the single row we are inserting.

Done

Details

Assignee

Reporter

Time tracking

1d 4h logged

Fix versions

Priority

Smart Checklist

Created September 15, 2022 at 6:46 AM
Updated May 8, 2024 at 5:56 AM
Resolved January 9, 2023 at 6:46 AM