-Added new LocalityCostFunction and LocalityCandidateGenerator that
cache localities of every region/rack combination and mappings of every
region to its most local server and to its most local rack.
-Made LocalityCostFunction incremental so that it only computes locality
based on most recent region moves/swaps, rather than recomputing the
locality of every region in the cluster at every iteration of the
balancer
-Changed locality cost function to reflect the ratio of:
(Current locality) / (Best locality possible given current cluster)
Signed-off-by: tedyu <yuzhihong@gmail.com>
After enabling test hbase.master.procedure.TestServerCrashProcedure#testRecoveryAndDoubleExecutionOnRsWithMeta, this bug is found in ServerCrashProcedure
Signed-off-by: Michael Stack <stack@apache.org>
In Standalone mode with local filesystem HBase logs Warning message:Failed to invoke 'unbuffer' method in class org.apache.hadoop.fs.FSDataInputStream
Signed-off-by: Umesh Agashe <uagashe@cloudera.com>
Signed-off-by: Sean Busbey <busbey@apache.org>
-Added new LocalityCostFunction and LocalityCandidateGenerator that
cache localities of every region/rack combination and mappings of every
region to its most local server and to its most local rack.
-Made LocalityCostFunction incremental so that it only computes locality
based on most recent region moves/swaps, rather than recomputing the
locality of every region in the cluster at every iteration of the
balancer
-Changed locality cost function to reflect the ratio of:
(Current locality) / (Best locality possible given current cluster)
Signed-off-by: Sean Busbey <busbey@apache.org>
Signed-off-by: Chia-Ping Tsai <chia7712@gmail.com>
Unit test (TestAssignmentManager) uses mock which always aggregates. So added trace level log message and verified manually on a single node cluster.
Signed-off-by: Michael Stack <stack@apache.org>
Calling closeRegion() directly on remote server is not supported post-AMv2. Calling unassign() on master
Signed-off-by: Michael Stack <stack@apache.org>
This issue adds comments and a sort so system tables are queued first
(which will ensure they go out first). This should be good enough
along w/ existing scheduling mechanisms to ensure system/meta get
assigned first.
Signed-off-by: Michael Stack <stack@apache.org>
Following AMv2 procedures are modified to override onSubmit(), onFinish() hooks provided by HBASE-17888 to do
metrics calculations when procedures are submitted and finshed:
* AssignProcedure
* UnassignProcedure
* MergeTableRegionProcedure
* SplitTableRegionProcedure
* ServerCrashProcedure
Following metrics is collected for each of the above procedure during lifetime of a process:
* Total number of requests submitted for a type of procedure
* Histogram of runtime in milliseconds for successfully completed procedures
* Total number of failed procedures
As we are moving away from Hadoop's metric2, hbase-metrics-api module is used for newly added metrics.
Modified existing tests to verify count of procedures.
Signed-off-by: Michael Stack <stack@apache.org>
Umesh Agashe
Ben-Epstein
Kahlil Oppenheimer
Sean Busbey
tedyu
张世彬10204932
Guanghao Zhang
David Harju
Ramkrishna
Andrew Purtell
QilinCao
Ashish Singhi
Pankaj Kumar
Josh Elser
Michael Stack
Vincent
Ashu Pachauri
Chia-Ping Tsai
Gary Helmling
Yi Liang
Alex Araujo
huzheng
Phil Yang
zhangduo