1. ix bugs in query layer.
2. remove some redundant code in BE
3. support specify multi helper node when starting FE
4. add proc 'cluster_load_statistic' to show load balance situation of Palo
1. rewrite metric module, add disk capacity metric
2. add Cluster load statistic proc to observer cluster load balance status
3. fix bug: show table status from db throw Nullpointer exception
4. performance: change push_down table size to 1024 in HashJoinNode
1. modify FE audit log format, for parsing it more convenient
2. fix bug: evaluate conjuncts assigned to SortNode in InlineViewRef
3. fix bug: invalid apache hdfs broker jar in apache_hdfs_broker_java_libraries.tar.gz
1. call done->Run when stream receiver is closed
2. change the name of frontend in bdbje, use arbitrary name instead of ip_port
3. change sync backend report to async, to avoid timeout when backend report
4. fix: buf in WrapperField is not equal to field_buf minus one
merge to 2d4cc9e1358c980b4f726e17d036639bc31127aa (#188)
contains:
first_value with PRECEDING LEFT and NON-PRECEDING RIGHT rewrite error and count* materialize SlotDescriptor
error when referring to the slot of the current query and subquery Simultaneously.
fix join and count(*) materialize SlotDescriptor error.
fix materialize scannode's conjuncts bug.
remove no used materialization work.
it have to evaluate orderby in subquery because we limit the number of rows returned by subquery.
the method of judging limit is wrong.
user info is missing when retrying to call load check.
It's wrong to pass aggregate function when it's param is not materialized.
InsertStmt does not pass the session param to observer.
In the case when 3 FE (1 master and 2 followers) start at same time,
We may catch com.sleepycat.je.rep.DatabasePreemptedException which said that
"Database xx has been forcibly closed in order to apply a replicated remove operation."
