Fritz Mueller Colin Slater 

(skipped as John Gates was not able to join the meeting)

 Status, progress:

• discussed at the previous meeting while John Gates was on vacation. Details in: Database Meeting 2023-01-04
• improved monitoring effort:
  • 
  • 
• chasing the bug:
  • 
  • Also, see a theory on what might be causing the problem at these threads:
    • https://lsstc.slack.com/archives/G2JPZ3GC8/p1673434073496049
    • https://lsstc.slack.com/archives/G2JPZ3GC8/p1673464304435489

What's next?

Fabrice Jammes is working on upgrading the ingest application to the latest version of the Ingest system's API. There is a problem reported in:

• 
• the investigation has moved to qserv-dev at IDF as it was discussed at the last group meeting
• Igor Gaponenko is participating in the investigation

Updates:

• Fabrice Jammes is not seeing any issues with the integration test after moving it to qserv-dev at IDF
• Igor Gaponenko thinks we should still pursue the investigation using test Qserv at the French Kubernetes provider where the CI problem seems to be reproducible
• Fritz Mueller suggested adding more logging into GHA-based CI to allow seeing extended reports from the Ingest system

Next steps:

• Fabrice Jammes will set up another test Qserv at the French Kubernetes provider and work with Igor Gaponenko to reproduce and investigate conditions triggering the problem. 

Igor Gaponenko here an update on this subject:

• an extended report on the study can be found at Experiments with partitioned InnoDB tables in the MySQL server version 8
• the new version of the report now included results for SELECT queries as well.
• the general consensus on the potential use of MySQL in Qserv:
  • consider switching to MySQL 8 in the Qserv czar database in order to speed up the director  index builder and Qserv results merger/processor
  • keep using MariaDB on workers (table size is the main concern for switching the engine from MyISAM to InnoDB)

Fritz Mueller proposed to evaluate the table/row compression of both table engines in MySQL and MariaDB. Goals:

• the efficiency of the compression (disk usage reduction)
• the performance of the compression and its effect of it on the overall performance of the Ingest system, including its relevant stages, such as ingesting data, table management operations (index creation)
• what impact the compressed tables/rows would have on the performance of the SELECT queries (and Qserv in general)
• findings will be reported in the report linked above