potiuk commented on issue #59445:
URL: https://github.com/apache/airflow/issues/59445#issuecomment-3714539655

   One of the reasons might also be not upgrading to latest version. 3.1.5 had 
some "N+1" dag fixes (too many sql queries generated on dag list page) . What 
can also slow down things are huge history in dag files that might slow down 
queries (`airflow db clean` might help).
   
   Maybe those are limited resources on your database, not 
vacuumed/cleaned/rebuilt indexes - the usual database maintenance that should 
happen regardless of what application is using it. You should take a loook at 
the monitoring of your database, looking at excessive queries, peaks in 
resource usage, strange errors - possibly looking at plans of queries generated 
by your database and looking for full-table scans. Generally when database 
statistics are not updated and cleaned up regularly, your table stats migh 
wildy differ from the actual data stored and database engine might use 
sub-optimal ways of planning the queries, whihc - for example -might result in 
loading all rows to memory instead of just loading indexes.
   
   Also we are just about (next week latest) to have RC of 3.1.6 which you 
might want to take for a spin before we release it (watch the devlist for 
announcement) and it has even more small fixes and optimisations in those UI/DB 
queries generated and it might be that this will help.
   
   Another thing you might take a look at - Fast API (when run in containers) 
recommends ONE worker per container and multiple containers - so comparing to 
flask it's way better to have several running api-servers - each in a separate 
container rather than several workers in a single api-server (as 
counter-intuitive it might seem).
   
   
   You can try it all - ideally one-by-one and see if things are improving 
after every step - it would be great to have feedback if some of that helped.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to