wjddn279 commented on issue #56641:
URL: https://github.com/apache/airflow/issues/56641#issuecomment-3460906655

   @alkismavridis 
   
   The memory limit needs to be increased. Since the scheduler spawns 32 worker 
subprocesses, there is a significant memory load. I have observed that the 
memory of each worker process gradually increases over time and then stabilizes 
at a certain level. (This is likely due to the parent process’s memory being 
copy-on-write, but I am still investigating the root cause.)
   
   I recommend setting the restart threshold to 5 GB. It would also be helpful 
to increase the memory limit and use smem to check whether the memory of each 
worker process continues to grow over time. By my measurements, there is almost 
no change in memory usage at the following levels.
   
   ```
     PID User     Command                         Swap      USS      PSS      
RSS
       1 airflow  /usr/bin/dumb-init -- /entr        0       84      110      
888
   12079 airflow  /bin/bash                          0     1736     1835     
3484
   1212609 airflow  /usr/bin/python3 /usr/bin/s        0    14144    14268    
16000
      54 airflow  airflow serve-logs                 0   101836   107213   
175100
      52 airflow  airflow worker -- LocalExec        0   125588   128036   
201980
      26 airflow  airflow worker -- LocalExec        0   125652   128097   
201996
      53 airflow  airflow worker -- LocalExec        0   125664   128112   
202056
      38 airflow  airflow worker -- LocalExec        0   125928   128369   
202288
      36 airflow  airflow worker -- LocalExec        0   126588   129008   
202788
      40 airflow  airflow worker -- LocalExec        0   126736   129162   
203028
      47 airflow  airflow worker -- LocalExec        0   126764   129190   
203064
      50 airflow  airflow worker -- LocalExec        0   126820   129245   
203112
      31 airflow  airflow worker -- LocalExec        0   126960   129385   
203240
      46 airflow  airflow worker -- LocalExec        0   127048   129473   
203336
      44 airflow  airflow worker -- LocalExec        0   127092   129516   
203376
      39 airflow  airflow worker -- LocalExec        0   127112   129535   
203380
      41 airflow  airflow worker -- LocalExec        0   127180   129602   
203440
      35 airflow  airflow worker -- LocalExec        0   127192   129615   
203460
      42 airflow  airflow worker -- LocalExec        0   127228   129639   
203312
      34 airflow  airflow worker -- LocalExec        0   127440   129861   
203684
      29 airflow  airflow worker -- LocalExec        0   127484   129905   
203728
      43 airflow  airflow worker -- LocalExec        0   127540   129960   
203772
      32 airflow  airflow worker -- LocalExec        0   127772   130190   
203976
      33 airflow  airflow worker -- LocalExec        0   127832   130251   
204040
      49 airflow  airflow worker -- LocalExec        0   127848   130268   
204060
      27 airflow  airflow worker -- LocalExec        0   127852   130270   
204020
      28 airflow  airflow worker -- LocalExec        0   127852   130270   
204016
      51 airflow  airflow worker -- LocalExec        0   127852   130272   
204076
      30 airflow  airflow worker -- LocalExec        0   127856   130274   
204040
      48 airflow  airflow worker -- LocalExec        0   127864   130284   
204076
      37 airflow  airflow worker -- LocalExec        0   127876   130295   
204076
      45 airflow  airflow worker -- LocalExec        0   127900   130318   
204068
       7 airflow  /usr/python/bin/python3.12         0   301160   313669   
398636
   ```


-- 
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