Kannan has commented on the revision "[jira][89-fb]
good stuff! initial round of comments on this change.
src/main/java/org/apache/hadoop/hbase/client/HConnectionManager.java:1700 function name: plural -> singular?
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:42 shared the same destination -> which share the same destination
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:95 simplify both line 92 & 95 to:
the list of puts which could not be queued
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:262 is there a possibility of a Concurrent Modification Exception since this map is not declared a concurrent map?
src/main/java/org/apache/hadoop/hbase/client/HConnectionManager.java:1700 documentation needed for this function.
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:183 is this lookup safe, when another thread might be inserting into this map? Doesn't this need to be a concurrent hash map?
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:211 the name of the param should have "put" somewhere and the fact that it is a "per-region server" queue size. Also, add some doc for the param here...
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:362 elapse -> elapsed
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:353 document param to avoid confusion (such as if this is a time or count based frequency) and if time what the units are, etc.
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:345 seems like this value can sometimes be inconsistent if you this happens right in between line 375 & 376.
src/test/java/org/apache/hadoop/hbase/client/TestHTableMultiplexer.java:67 isSucceeded -> success
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:394 each failed puts -> each failed put
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:426 s/and there are/and
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:424 Processed put <N> requests --> Processed <N> put requests
src/main/java/org/apache/hadoop/hbase/client/HTableMultiplexer.java:387 Is there a way to avoid preparing this map? At the very least, we can be lazy about creating this map when a failure actually happens. If there are 0 failures, we don't use this map.
What comparator will be used for the "Put" keys in this map?
To: Kannan, Liyin
Cc: JIRA, tedyu