Ngone51 commented on pull request #31876:
URL: https://github.com/apache/spark/pull/31876#issuecomment-822550411


   > This is making assumption about the type - while it would work currently 
with existing implementation, how is this going to work going forward ?
   
   > Yeah, in this case, we need to deal with those 
asInstanceOf[BlockManagerId] type cast to make it not fail for other location 
type. Any thoughts there?
   
   @hiboyang @mridulm Answer your both questions together since they are the 
same. 
   
   You can see I've added two interfaces (`ExecutorLocation`, `HostLocation`) 
if you look at the latest change. Thus, we'd match `ExecutorLocation` or 
`HostLocation` where we need executor id or host and default to `None` if the 
location doesn't extend `ExecutorLocation` / `HostLocation`. As a result, we 
won't fail if the location is not a `BlockManagerId`.
   
   That also means we'd skip some operations for locations that not based on 
executor or host, e.g.,  unregister mapstatus on executor lost, update 
healthracker,  because these concepts (executorId, host) are no longer suitable 
for those locations. 
   
   
   


-- 
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]



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

Reply via email to