Revert "Revert "HBASE-2004 TestClientClusterStatus is flakey""
This is a revert of a revert, i.e. a reapplication, just so I can fix
the JIRA number.
This reverts commit 6796b8e21f.
LocalHBaseCluster forces random port assignment for sake of concurrent unit test
execution friendliness, but we still need a positive test for RPC and info port
assignment.
The 1.x functionality of Master DDL operations is that "post" observer hooks
are only invoked when the DDL action was successful. With the async-ness of
ProcV2, we find ourselves in a case where the post-hook may be invoked before
the Procedure runs and fails. We need to introduce some blocking to wait and
see if the Procedure is going to fail on a precondition before invoking the hook.
Signed-off-by: Michael Stack <stack@apache.org>
Remove assert in splittableregionprocedure. It was in the prepare.
Was causing fail in legit case where a region split follows a
table split BEFORE the parent has been GC'd. The region split
finds the parent in SPLIT state which is right. The assert was
having us fail. No need.
Also disabled TestHTrace since not supported in 2.0.0 and flakey.
Only call server.checkIfShouldMoveSystemRegionAsync if a node has been
added. Do not call it if only one regionserver in cluster. Make it
so ServerCrashProcedure runs before it. Add logging if
server.checkIfShouldMoveSystemRegionAsync was responsible for
MOVE (Previous was a mystery when it cut in).
Previous we'd call it when there was a nodeChildrenChanged. These
happen before nodeDeleted. If a server crashed,
checkIfShouldMoveSystemRegionAsync could run first, find the
server that had not yet registered as crashed, find system
tables on it and then try to move them. It would fail because
server would not respond to RPC. The region move would then
be waiting on the servercrashprocedure to wake it up when
done processing but this move had locked the region so
SCP couldn't run....
Serializing, if appropriate, write the hbase-1.x version of the
Comparator to the hfile trailer so hbase-1.x files can read hbase-2.x
hfiles (they are the same format).
This patch adds mirroring of table state out to zookeeper. HBase-1.x
clients look for table state in zookeeper, not in hbase:meta where
hbase-2.x maintains table state.
The patch also moves and refactors the 'migration' code that was put in
place by HBASE-13032.
D hbase-client/src/main/java/org/apache/hadoop/hbase/CoordinatedStateException.java
Unused.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
Move table state migration code from Master startup out to
TableStateManager where it belongs. Also start
MirroringTableStateManager dependent on config.
A hbase-server/src/main/java/org/apache/hadoop/hbase/master/MirroringTableStateManager.java
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/TableStateManager.java
Move migration from zookeeper of table state in here. Also plumb in
mechanism so subclass can get a chance to look at table state as we do
the startup fixup full-table scan of meta.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/AssignmentManager.java
Bug-fix. Now we create regions in CLOSED state but we fail to check
table state; were presuming table always enabled. Meant on startup
there'd be an unassigned region that never got assigned.
A hbase-server/src/test/java/org/apache/hadoop/hbase/master/TestMirroringTableStateManager.java
Test migration and mirroring.
Value of 'htd' is null as it is initialized in the constructor but when the object is deserialized its null. Got rid of member variable htd and made it local to method.
* Added a comment in MergeTableRegionsProcedure and SplitTableRegionProcedure explaining specific rollbacks has side effect that AssignProcedure/s are submitted asynchronously and those procedures may continue to execute even after rollback() is done.
* Updated comments in tests with correct rollback state to abort
* Added overloaded method MasterProcedureTestingUtility#testRollbackAndDoubleExecution which takes additional argument for waiting on all procedures to finish before asserting conditions
* Updated TestMergeTableRegionsProcedure#testRollbackAndDoubleExecution and TestSplitTableRegionProcedure#testRollbackAndDoubleExecution to use newly added method
Signed-off-by: Michael Stack <stack@apache.org>
Fix two issues:
# Meta Replicas can all be assigned to the same server. This
will call the test to hang when we do our kill of the server
hosting meta because there'll be no replicas to read from
as test intends. Check is to look for this condition on
startup and adjust if we come across it. Replicas cross-cut
assignment. They need work.
# Other issue was shutdown. The master started toward the
end of the test may not have come up fully by the time
shutdown is called. We could be stuck assigning the
meta replicas. Have shutdown shutdown the procedure
executor engine.
There is other cleanup and notes in the below.
M HMaster
Remove the silly stops in startup now we have real
means of shutting down Master during init.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/MasterMetaBootstrap.java
This replica stuff was doing stuff it shouldn't be doing
like setting core Master state flags. It may have made
sense once but now meta is assigned by a Pv2 Procedure
so the flag setting in here is meddlesome. Clear out
methods no longer needed.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/AssignmentManager.java
Remove unused methods.
Changes local variable names so they align w/ our naming elsewhere in
code base.
M hbase-server/src/test/java/org/apache/hadoop/hbase/client/TestMetaWithReplicas.java
Check for all replicas on the one server.
On Master@shutdown, close the shared Master connection to kill any
ongoing RPCs by hosted clients.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
Call close ont the Master shared clusterconnection to kill any ongoing
rpcs.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/ServerManager.java
Remove guts of close; we were closing the Masters connection....not
our responsibility.
Added unit test written by Duo Zhang which demonstrates the case where
Master will not go down.
Signed-off-by: zhangduo <zhangduo@apache.org>
Kill backup master first
Add some cleanup around NamespaceManager
Shorten the timeout waiting on namespace manager as workaround
until we have better soln for interrupting ongoing client rpcs.
Do it in general for all tests.
Signed-off-by: zhangduo <zhangduo@apache.org>
Rename the PE Worker threads.
Send an interrupt if worker taking a long time to go down
(it may be RPC'ing out to a dead server, retrying so
interrupt). Also join on the ProcedureExecutor shutting down.
This will make problems shutting down more obvious.
Disable TestRegionsOnMasterOptions. Master carrying Regions is broke.
Set the ProcedureExcecutor worker threads as daemon.
Ditto for the timeout thread.
Remove hack from TestRegionsOnMasterOptions that was
put in place because the test would not go down.
Michael Stack
Josh Elser
zhangduo
Esteban Gutierrez
Andrew Purtell
tedyu
Mike Drob
anastas
Chia-Ping Tsai
Reid Chan
eshcar
Umesh Agashe
Thiruvel Thirumoolan
Ben Lau
Toshihiro Suzuki
Pankaj Kumar
Vasudevan
Apekshit Sharma
huzheng
Guanghao Zhang
Aman Poonia
Balazs Meszaros
Rajeshbabu Chintaguntla
Ashish Singhi
Peter Somogyi
kewang
Rahul Gidwani
binlijin
Duo Zhang
Jan Hentschel
haxiaolin
Sergey Soldatov
Artem Ervits
Guangxu Cheng
lixiang