mxm commented on issue #15031:
URL: https://github.com/apache/iceberg/issues/15031#issuecomment-3743123485

   Thanks for raising the issue @QiuYucheng2003! Just adding a bit more Flink 
context to what's already been observed.
   
   Flink has two distinct deployment modes:
   
   1. Application clusters - only spun up for one particular job and shut down 
when the job completes / gets canceled. 
   2. Session clusters - long-lived clusters, which can host any number of jobs.
   
   Flink recycles the classloader within a given job, even when the job 
restarts. So we should not see this problem in application clusters because 
there will only ever be one job. Session clusters, on the other hand, create a 
new classloader for every job and hence accumulate thread pools across these 
jobs, until the cluster gets shut down. 
   
   Generally, using static variables which reference user classes is strongly 
discouraged for Flink session clusters. It looks like we need an explicit 
shutdown call which we can run when Flink jobs terminate.


-- 
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]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to