szlta commented on a change in pull request #909: HIVE-22821
URL: https://github.com/apache/hive/pull/909#discussion_r389958291
##########
File path:
llap-server/src/java/org/apache/hadoop/hive/llap/io/api/impl/LlapIoImpl.java
##########
@@ -221,6 +227,9 @@ public void debugDumpShort(StringBuilder sb) {
metadataCache, dataCache, bufferManagerOrc, conf, cacheMetrics,
ioMetrics, tracePool);
this.genericCvp = isEncodeEnabled ? new GenericColumnVectorProducer(
serdeCache, bufferManagerGeneric, conf, cacheMetrics, ioMetrics,
tracePool) : null;
+ proactiveEvictionExecutor = Executors.newSingleThreadExecutor(
Review comment:
Okay, so suppose you use beeline and do a drop table statement:
A thread from HiveServer2-Handler-Pool will pick up the incoming Thrift
request and execute.
Execution will go into DropTableOperation. Anything happening here will
prevent this thread from finishing and thus you as the user on the beeline side
(as thrift connects you) will have to wait.
If we invoke a proactive eviction from DropTableOperation (which this change
does) obviously the user on beeline side will see this addition processing time
for their query.
Now if this additional time is long - that could be bad, since they only
wanted to drop a table and now they're waiting extra seconds if the
DropTableOperation 'waits' for many LLAP daemons to finish their eviction on
their big cache structure.
----------------------------------------------------------------
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]
