Skip to:

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

Write generic script to extract SQL from pstress logs.

Description

Extract SQLs from default sql logs file  from pstress logs and execute/save them with same timestamp as executed in default sql logs.

for example  default_1.sql 12 SQL1 14 SQL2 default_2.sql 10 SQL3 12 SQL4 13 SQL5 open two different session. and in one session at time second 2, 4 execute SQL1,2. in second session at time second 0,2,3 execute SQL3,4,5

To make things simple. create separate session and execute them based on the session time. no coordination between thread is required.

Modify pstress to save the time format as required by this script. 

Environment

None

Smart Checklist

Activity

mohit.joshi May 5, 2022 at 3:14 AM
Edited

Hi  - 

There are 2 things here:

  1. Writing a shell script that converts pstress thread logs into an executable SQL file.

  2. Executing the SQLs in the time order pstress executed them

Point 1 - It is possible to write a script that can convert pstress thread logs into SQL files which can be directly executed on a running server. 

Point 2 - This can be achieved in case of single-threaded runs. It will become too complicated to handle, when the crash/bug is found in nth trial and/or X no. of threads.

1. Firstly, we will need a running server before the crash happened to match the DB state which means, the scope of this script will have to involve starting,stopping server from incident directory.
2. Secondly, if X no. of threads, we will need to open X number of client sessions and then somehow based on timestamps run some queries from file1, file2 and so on.
IMO, this is something we should do by improving seed functionality in pstress and build an option like --replay which can start the DB server prior to crash and then using same seed try to generate and execute queries.

So, I think we should focus on writing a script that does Point no. 1. It will be a handy script that will quickly generate the SQL file. The QA user can execute it against a running server to repeat the crash

Won't Do

Details

Assignee

Reporter

Time tracking

1d 1h 30m logged

Priority

Smart Checklist

Created May 4, 2022 at 7:52 AM
Updated March 6, 2024 at 5:53 PM
Resolved December 8, 2022 at 4:16 AM

Flag notifications