Rajshekhar Muchandi created LIVY-996:
----------------------------------------

             Summary: Configure interval check of session garbage
                 Key: LIVY-996
                 URL: https://issues.apache.org/jira/browse/LIVY-996
             Project: Livy
          Issue Type: Improvement
          Components: Server
            Reporter: Rajshekhar Muchandi
            Assignee: Rajshekhar Muchandi


We observed multiple deletion of the same session because a deletion of a 
session could take more than a minute for various reasons: session not existing 
anymore (especially true when restarting), session time to stop takes time, 
kube api slowness, timeouts etc...

In case a session takes more than a minute to stop, the collectGarbage() 
function, which is launched every minute, will relaunch the deletion of the 
session.

Moreover, each deletion is calling a future, hence a new thread pops up, in a 
cluster with 100s or 1000s retained, that could end up to high number of 
threads. And if these threads are limited and queued for operations in a thread 
pool, for sure these operations could take more than a minute.

To mitigate this effect and avoid warnings and relaunch of session's deletion 
multiple times, that would be nice to be able to configure this interval check 
session finished in state store instead of having it forced at 1 minute.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to