HeartSaVioR commented on issue #25577: [WIP][CORE][SPARK-28867] InMemoryStore 
checkpoint to speed up replay log file in HistoryServer
URL: https://github.com/apache/spark/pull/25577#issuecomment-534003329
 
 
   > Oh, sorry. I was just discussing the possible integrate way in the 
following design and not ask you to make the change right now.
   
   Yeah. I feel I was over-reacted. Sorry about that. Just wanted to not going 
back and spend another time to wait for new thing.
   
   > Sorry, I don't know this part well. Why do we need to implement sync and 
async dump separately ?
   
   It's not always easy to implement async dump for all possible 
implementations of KVStore (though KVStore is private API so for now we only 
have InMemoryStore and LevelDB...). We only discussed an idea of async snapshot 
for InMemoryStore - LevelDB is not considered here. So for now even in happy 
case InMemoryStore will support asynchronous dump and LevelDB will not.
   
   Suppose a case, LevelDB is provided to the AppStatusListener as underlying 
KVStore. (That wouldn't be true for live application but let's see it as 
interface level of view - as it receives ElementTrackingStore.) If you call 
dump() it should run as synchronous, and listener would be stuck until dump() 
has been finished and open the chance to drop events due to heavy latency. 
That's what we want to avoid and the reason to discuss asynchronous approach, 
right? For live application of AppStatusListener, underlying KVStore must 
support asynchronous dump. Otherwise it should just fail or give up dumping 
KVStore which caller should indicate and make decision.

----------------------------------------------------------------
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

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

Reply via email to