tillrohrmann commented on issue #9552: [FLINK-13884] Set default failure rate 
restart strategy delay to 0s
URL: https://github.com/apache/flink/pull/9552#issuecomment-527064874
 
 
   > > The restart storms should no longer happen with queued scheduling 
because we wait for the slot request timeout before failing again.
   > 
   > Hi @tillrohrmann , we once experienced restart storms in which case a task 
failed fast once it is deployed to TM, e.g. failed due to task internal errors 
or failed when trying to connecting to a bad external node. Queued scheduling 
might not help in such cases.
   
   You are right @zhuzhurk. Restart storms caused by external system overload 
won't be solved with the slot request timeouts.
   
   Then let's see this PR as a unification of the restart delays of the fixed 
delay and failure rate restart strategies. Whether to increase the default 
delay or not will be dealt with in a follow up issue. I've started a [survey on 
the 
ML](https://lists.apache.org/thread.html/107b15de6b8ac849610d99c4754715d2a8a2f32ddfe9f8da02af2ccc@%3Cdev.flink.apache.org%3E).
 Depending on the outcome of it, I'll create a new issue to update the delay. 

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


With regards,
Apache Git Services

Reply via email to