The archived Procedure WALs are moved to <hbase_root>/oldWALs/masterProcedureWALs
directory. TimeToLiveProcedureWALCleaner class was added which
regularly cleans the Procedure WAL files from there.
The TimeToLiveProcedureWALCleaner is now added to
hbase.master.logcleaner.plugins to clean the 2 WALs in one run.
A new config parameter is added hbase.master.procedurewalcleaner.ttl
which specifies how long a Procedure WAL should stay in the
archive directory.
Signed-off-by: Michael Stack <stack@apache.org>
This reverts commit 0d0c330401.
Backing out filterlist regression, see HBASE-18957. Work continuing branch for HBASE-18410.
Signed-off-by: Peter Somogyi <psomogyi@cloudera.com>
Signed-off-by: Michael Stack <stack@apache.org>
This reverts commit df34300cd3.
Backing out filterlist regression, see HBASE-18957. Work continuing branch for HBASE-18410.
Signed-off-by: Peter Somogyi <psomogyi@cloudera.com>
Signed-off-by: Michael Stack <stack@apache.org>
This reverts commit f54cc1ca51.
Backing out filterlist regression, see HBASE-18957. Work continuing branch for HBASE-18410.
Signed-off-by: Peter Somogyi <psomogyi@cloudera.com>
Signed-off-by: Michael Stack <stack@apache.org>
Patch ArrayIndexOutOfBoundsException when current() is called after
advance() has already returned false
Signed-off-by: Chia-Ping Tsai <chia7712@gmail.com>
These functions have been changed to return Optional<T> instead of T, where T = old return type.
- ObserverContext#getCaller
- RpcCallContext#getRequestUser
- RpcCallContext#getRequestUserName
- RpcServer#getCurrentCall
- RpcServer#getRequestUser
- RpcServer#getRequestUserName
- RpcServer#getRemoteAddress
- ServerCall#getRequestUser
Change-Id: Ib7b4e6be637283755f55755dd4c5124729f7052e
Signed-off-by: Apekshit Sharma <appy@apache.org>
CompactionRequest was removed from CP in HBASE-18453, this change reintroduces
CompatcionRequest to CP as a read-only interface called CompactionRequest.
The CompactionRequest class is renamed to CompactionRequestImpl.
Additionally, this change removes selectionTimeInNanos from CompactionRequest and
uses selectionTime as a replacement. This means that CompactionRequest:toString
is modified and compare as well.
Signed-off-by: Michael Stack <stack@apache.org>
- Change Service Coprocessor#getService() to List<Service> Coprocessor#getServices()
- Checkin the finalized design doc into repo
- Added example to javadoc of Coprocessor base interface on how to implement one in the new design
------------------------------------------------------
TL;DR
------------------------------------------------------
We are moving from Inheritence
- Observer *is* Coprocessor
- FooService *is* CoprocessorService
To Composition
- Coprocessor *has* Observer
- Coprocessor *has* Service
------------------------------------------------------
Design Changes
------------------------------------------------------
- Adds four new interfaces - MasterCoprocessor, RegionCoprocessor, RegionServierCoprocessor,
WALCoprocessor
- These new *Coprocessor interfaces have a get*Observer() function for each observer type
supported by them.
- Added Coprocessor#getService() to base interface. All extending *Coprocessor interfaces will
get it from the base interface.
- Added BulkLoadObserver hooks to RegionCoprocessorHost instad of SecureBulkLoadManager doing its
own trickery.
- CoprocessorHost#find*() fuctions: Too many testing hooks digging into CP internals.
Deleted if can, else marked @VisibleForTesting.
------------------------------------------------------
Backward Compatibility
------------------------------------------------------
- Old coprocessors implementing *Observer won't get loaded (no backward compatibility guarantees).
- Third party coprocessors only implementing Coprocessor will not get loaded (just like Observers).
- Old coprocessors implementing CoprocessorService (for master/region host)
/SingletonCoprocessorService (for RegionServer host) will continue to work with 2.0.
- Added test to ensure backward compatibility of CoprocessorService/SingletonCoprocessorService
- Note that if a coprocessor implements both observer and service in same class, its service
component will continue to work but it's observer component won't work.
------------------------------------------------------
Notes
------------------------------------------------------
Did a side-by-side comparison of CPs in master and after patch. These coprocessors which were just
CoprocessorService earlier, needed a home in some coprocessor in new design. For most it was clear
since they were using a particular type of environment. Some were tricky.
- JMXListener - MasterCoprocessor and RSCoprocessor (because jmx listener makes sense for
processes?)
- RSGroupAdminEndpoint --> MasterCP
- VisibilityController -> MasterCP and RegionCP
These were converted to RegionCoprocessor because they were using RegionCoprocessorEnvironment
which can only come from a RegionCPHost.
- AggregateImplementation
- BaseRowProcessorEndpoint
- BulkDeleteEndpoint
- Export
- RefreshHFilesEndpoint
- RowCountEndpoint
- MultiRowMutationEndpoint
- SecureBulkLoadEndpoint
- TokenProvider
Change-Id: I813145f2bc11815f52ac703563b879962c249764
Adjust exception control flow to fix findbugs warning
NP_NULL_ON_SOME_PATH_EXCEPTION, Possible null pointer dereference of
regionSink in org.apache.hadoop.hbase.tool.Canary$RegionMonitor.run()
on exception path
Added assertion checks to make sure that the error code for the
ToolRunner run() method is used.
Testing Done: Checked that TestCanaryTool unit tests fail when there is
an error code in the current Canary run.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
Changed the type hierarchy of Canary sinks to reduce confusion and avoid
cast errors.
Testing Done: Ran the TestCanaryTool.java test suite and confirmed that
the working is correct.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
This provides significant cluster start time reduction for FileSystems which do not surface locality (S3).
Signed-off-by: Andrew Purtell <apurtell@apache.org>
Upgrade commons-math:2.2 to commons-math3:3.6.1
Remove commons-math 2 specific content from LICENSE.vm
Add missing jersey-client dependency to hbase-it module
Signed-off-by: Michael Stack <stack@apache.org>
This is based on patch sent me by Balazs Meszaros. The good stuff in
here is from him. This patch does less than his ambition. It changes
Admin class only. Can work on making AsyncAdmin cohere in a follow-on.
* Deprecates getAlterStatus. Everywhere else we talk of 'modify' rather
'alter' and should use Future returned from async instead.
* isTableAvailable(TableName, byte [][]) has been deprecated to be
removed; use the overrie instead. This is a weird method.
* Changed listTableDescriptor to getDescriptor.
* Renamed other like methods to have same pattern (deprecating the old):
balancer => balance
setBalancerRunning => balancerSwitch
setNormalizerRunning => normalizerSwitch
enableCatalogJanitor => catalogJanitorSwitch
setCleanerChoreRunning => cleanerChoreSwitch
setSplitOrMergeEnabled => splitOrMergeEnabledSwitch
* Renamed (with deprecation of old) runCatalogScan => runCatalogJanitor.
* Reviewed generated javadoc and made some edits; purged reference to
hbase issues from our API, fixed param names, etc.
* Made all the enable services methods have same pattern.
* Renamed takeSnapshotAsync as snapshotAsync (with deprecation of old)
* Renamed execProcedureWithRet as execProcedureWithReturn (with
deprecation)
Signed-off-by: Michael Stack <stack@apache.org>
Javadoc for following methods are updated:
* Table.put(List<Put> puts)
* Table.delete(List<Delete> deletes)
Added @apiNote for delete regarding input list will not be modied in version 3.0.0
Signed-off-by: Michael Stack <stack@apache.org>
Main changes:
- ProcedureInfo and LockInfo were removed, we use JSON instead of them
- Procedure and LockedResource are their server side equivalent
- Procedure protobuf state_data became obsolate, it is only kept for
reading previously written WAL
- Procedure protobuf contains a state_message field, which stores the internal
state messages (Any type instead of bytes)
- Procedure.serializeStateData and deserializeStateData were changed slightly
- Procedures internal states are available on client side
- Procedures are displayed on web UI and in shell in the following jruby format:
{ ID => '1', PARENT_ID = '-1', PARAMETERS => [ ..extra state information.. ] }
Signed-off-by: Michael Stack <stack@apache.org>
* testSimpleMasterFailover - fixed and verified
* testPendingOpenOrCloseWhenMasterFailover - removed as logic is based on old code and no longer relevant. TestServerCrashProcedure tests assignments with crashing master and region servers
* testMetaInTransitionWhenMasterFailover - verified that it is fixed by patch for HBASE-18511.
Signed-off-by: Michael Stack <stack@apache.org>
Upgrade commons-collections:3.2.2 to commons-collections4:4.1
Add missing dependency for hbase-procedure, hbase-thrift
Replace CircularFifoBuffer with CircularFifoQueue in WALProcedureStore and TaskMonitor
Signed-off-by: Sean Busbey <busbey@apache.org>
Signed-off-by: Chia-Ping Tsai <chia7712@gmail.com>
This correctly casts the operand to a long to avoid negative offsets created by sign extending the integer operand.
Signed-off-by: tedyu <yuzhihong@gmail.com>
Do a pass with dependency:analyze; remove unused and
explicity list the dependencies we exploit.
Remove the parent dependencies set which had junit, mockito,
log4j, and findbugs annotations (had to put junit back
temporarily in subsequent version of this patch TODO). Listing in
parent set meant these libs were dependencies for all modules
which in practice was not the case. Edited all modules so
those that need any from this parent set now do explicit listing.
Ran the dependency:analyze over the project. Acted on most
suggested removals and requests for explicit listing. Some
grey areas remain around transitives that come in with
hadoop -needs better excludes, another project- and that
the dependency:analyze tool is not always accurate in its
reporting.
Move Driver to be the main-class in hbase-mapreduce jar rather than
in the hbase-server jar.
Reference the hbase-server and shaded protobuf so they get bundled
when you do 'hbase mapredcp'.
classpath issue when running as a developer fixed
removed thrift webapp from server (it's not used at all, since moved to thrift webapp)
Signed-off-by: tedyu <yuzhihong@gmail.com>
- Moves out o.a.h.h.{mapred, mapreduce} to new hbase-mapreduce module which depends
on hbase-server because of classes like *Snapshot{Input,Output}Format.java, WALs, replication, etc
- hbase-backup depends on it for WALPlayer and MR job stuff
- A bunch of tools needed to be pulled into hbase-mapreduce becuase of their dependencies on MR.
These are: CompactionTool, LoadTestTool, PerformanceEvaluation, ExportSnapshot
This is better place of them than hbase-server. But ideal place would be in separate hbase-tools module.
- There were some tests in hbase-server which were digging into these tools for static util funtions or
confs. Moved these to better/easily shared place. For eg. security related stuff to HBaseKerberosUtils.
- Note that hbase-mapreduce has secondPartExecution tests. On my machine they took like 20 min, so maybe
more on apache jenkins. That's basically equal reduction of runtime of hbase-server tests, which is a
big win!
Change-Id: Ieeb7235014717ca83ee5cb13b2a27fddfa6838e8
Breaking change to our ReplicationEndpoint and BaseReplicationEndpoint.
ReplicationEndpoint implemented Guava 0.12 Service. An abstract
subclass, BaseReplicationEndpoint, provided default implementations
and facility, among other things, by extending Guava
AbstractService class.
Both of these HBase classes were marked LimitedPrivate for
REPLICATION so these classes were semi-public and made it so
Guava 0.12 was part of our API.
Having Guava in our API was a mistake. It anchors us and the
implementation of the Interface to Guava 0.12. This is untenable
given Guava changes and that the Service Interface in particular
has had extensive revamp and improvement done. We can't hold to
the Guava Interface. It changed. We can't stay on Guava 0.12;
implementors and others on our CLASSPATH won't abide being stuck
on an old Guava.
So this class makes breaking changes. The unhitching of our Interface
from Guava could only be done in a breaking manner. It undoes the
LimitedPrivate on BaseReplicationEndpoint while keeping it for the RE
Interface. It means consumers will have to copy/paste the
AbstractService-based BRE into their own codebase also supplying their
own Guava; HBase no longer 'supplies' this (our Guava usage has
been internalized, relocated).
This patch then adds into RE the basic methods RE needs of the old
Guava Service rather than return a Service to start/stop only to go
back to the RE instance to do actual work. A few method names had to
be changed so could make implementations with Guava Service internally
and not have RE method names and types clash). Semantics remained the
same otherwise. For example startAsync and stopAsync in Guava are start
and stop in RE.
* Fixed ServerCrashProcedure to set forceNewPlan to false for instances AssignProcedure. This enables balancer to find most suitable target server
* Fixed and enabled TestRestartCluster#testRetainAssignmentOnRestart on master
* Renamed method ServerName@isSameHostnameAndPort() to isSameAddress()
Signed-off-by: Michael Stack <stack@apache.org>
Instead of using an Atomic Reference to data and aborting when we detect
that new data comes in, use the native cancellation/pre-emption features
of Java Future.
Behavior prior to these changes is to call expireServer(), log exception and suppress it. These changes will result in RS receiving the YouAreDeadException and treating it as a fatal error. This 'fail fast' approach will help us stabilize the code. This behavior can be reconsidered later if necessary.
Signed-off-by: Michael Stack <stack@apache.org>
Changes the configuration hbase.balancer.tablesOnMaster from list of
table names to instead be a boolean; true if master carries
tables/regions and false if it does not.
Adds a new configuration hbase.balancer.tablesOnMaster.systemTablesOnly.
If true, hbase.balancer.tablesOnMaster is considered true but only
system tables are put on the master.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/HMaster.java
Master was claiming itself active master though it had stopped. Fix
the activeMaster flag. Set it to false on exit.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/LoadBalancer.java
Add new configs and convenience methods for getting current state of
settings.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
Move configs up into super Interface and now the settings mean
different, remove the no longer needed processing.
* calls to methods getLowestLocalityRegionServer() & getLeastLoadedTopServerForRegion() got removed in HBASE-18164
* call to calculateRegionServerLocalities() got removed in HBASE-15486
* Some other minor improvements
Change-Id: Ib149530d8d20c019b0891c026e23180e260f59db
Signed-off-by: Apekshit Sharma <appy@apache.org>
Before this commit, BucketCache always used the default values.
This commit adds the ability to configure these values.
Signed-off-by: tedyu <yuzhihong@gmail.com>
Add in hbase-thirdparty hbase-shaded-netty instead.
s/io.netty/org.apache.hadoop.hbase.shaded.io.netty/ everywhere in hbase.
Also set a system property when running tests and when starting
hbase; required by netty so can find the relocation files in the
bundled .so.
If an unassign is unable to communicate with its target server,
expire the server and then wait on a signal from ServerCrashProcedure
before proceeding. The unassign has lock on the region so no one else
can proceed till we complete. We prevent any subsequent assign from
running until logs have been split for crashed server.
In AssignProcedure, do not assign if table is DISABLING or DISABLED.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/RegionTransitionProcedure.java
Change remoteCallFailed so it returns boolean on whether implementor
wants to stay suspended or not.
M hbase-server/src/main/java/org/apache/hadoop/hbase/master/assignment/UnassignProcedure.java
Doc. Also, if we are unable to talk to remote server, expire it and
then wait on SCP to wake us up after it has processed logs for failed
server.
zhangduo
tedyu
Ramkrishna
Jan Hentschel
Peter Somogyi
anoopsamjohn
Sean Busbey
Chia-Ping Tsai
Michael Stack
Guangxu Cheng
Maytee Chinavanichkit
Guanghao Zhang
Apekshit Sharma
libisthanks
libisthanks
Ashu Pachauri
Andrew Purtell
Andy Yang
Pankaj Kumar
huzheng
Chinmay Kulkarni
Josh Elser
Zach York
Abhishek Singh Chouhan
Yechao Chen
brandboat
Reid Chan
Umesh Agashe
Yun Zhao
Biju Nair
Balazs Meszaros
Michael Stack
Cesar Delgado
Sean Mackrory
Tamas Penzes
Vikas Vishwakarma
Chun-Hao Tang
Ajay Jadhav
Huaxiang Sun
Chia-Ping Tsai
Vladimir Rodionov
Mike Drob
anastas
Ashish Singhi
Vladimir Rodionov
Lars Francke
Deon Huang
Jerry He
Yi Liang
Esteban Gutierrez
Yu Li