Created a new WALKey Interface and a WALKeyImpl. The WALKey Interface
is surfaced to Coprocessors and throughout most of the code base.
WALKeyImpl is used internally by WAL and by Replication which need
access to WALKey setters.
Methods that were deprecated in WALObserver because they were exposing
Private audience Classes have been undeprecated now we have WALKey.
Moved over to use SequenceId#getSequenceId throughout. Changed
SequenceId#getSequenceId removing the IOE.
- Changed testThreeRSAbort to kill the RSs intead of aborting. Simple aborting will close the regions, we want extreme failure testing here.
- Adds some logging for easier debugging.
- Refactors TestDistributedLogSplitting to use standard junit rules.
Move the hadoop-hdfs guava exclude in modules up to the top pom.
Looks like an exclude in a module is not additive but rather exclusive
blanking out the top level set of exclusions.
Tested by looking in lib dir of the built tarball.
RowCounter and other related HBase's MapReduce classes have been moved
to hbase-mapreduce component by HBASE-18640, related chapter was
out-of-date and this fix replaced hbase-server with hbase-mapreduce
to correct those commands
Also this change moved RowCounter_Counters.properties to
hbase-mapreduce package as well
JIRA https://issues.apache.org/jira/browse/HBASE-19023
Signed-off-by: tedyu <yuzhihong@gmail.com>
null from other coprocessor even on bypass
If 'bypass' is set by a Coprocessor, skip out on calling any subsequent
Coprocessors that might be chained to a bypassable method.
This patch restores some of the 'complete' behavior removed by
HBASE-19123 only 'bypass' now triggers 'complete'.
For a egionserver's view of a table (the regions
that belong to a table hosted on a regionserver),
this change tracks the latencies of operations that
affect the regions for this table.
Tracking at the per-table level avoids the memory bloat
and performance impact that accompanied the previous
per-region latency metrics while still providing important
details for operators to consume.
Signed-Off-By: Andrew Purtell <apurtell@apache.org>
- Adding javadoc comments
- Bug: ServerStateNode#regions is HashSet but there's no synchronization to prevent concurrent addRegion/removeRegion. Let's use concurrent set instead.
- Use getRegionsInTransitionCount() directly to avoid instead of getRegionsInTransition().size() because the latter copies everything into a new array - what a waste for just the size.
- There's mixed use of getRegionNode and getRegionStateNode for same return type - RegionStateNode. Changing everything to getRegionStateNode. Similarly rename other *RegionNode() fns to *RegionStateNode().
- RegionStateNode#transitionState() return value is useless since it always returns it's first param.
- Other minor improvements
Update timeouts for TestRegionObserverInterface.
Reason: There are ~10 tests there, each with 5 min individual timeout. Too much. The test class is labelled MediumTests,
let's used that with our standard CategoryBasedTimeout. 3 min per test function should be enough even on slower Apache machines.
This avoids the situation where the build machine has sufficient disk
space (a few GB's at most) to run an HBase test, but the default YARN
configuration would preclude the NM's from starting correctly. This
should eliminate a trivial source of build flakiness based on the host
machines being used.
Signed-off-by: Ted Yu <tedyu@apache.org>
Removed TestHRegion#testMemstoreSizeWithFlushCanceling test.
CPs are not able to cancel a flush any more so this test was
failing; removed it (It was testing memory accounting kept
working across a cancelled flush).
Signed-off-by: Michael Stack <stack@apache.org>
Use header and footer in our *.jsp pages to avoid unnecessary redundancy (copy-paste of code)
Misc edits:
- Due to redundancy, new additions make it to some places but not others. For eg there are missing links to "/logLevel", "/processRS.jsp" in few places.
- Fix processMaster.jsp wrongly pointing to rs-status instead of master-status (probably due to copy paste from processRS.jsp)
- Deleted a bunch of extraneous "</a>" in processMaster.jsp & processRS.jsp
- Added missing </div> tag in snapshot.jsp
- Deleted fossils of html5shiv.js. It's uses and the js itself were deleted in the commit "819aed4ccd073d818bfef5931ec8d248bfae5f1f"
- Fixed wrongly matched heading tags
- Deleted some unused variables
Tested:
Ran standalone cluster and opened each page to make sure it looked right.
Sidenote:
Looks like HBASE-3835 started the work of converting from jsp to jamon, but the work didn't finish. Now we have a mix of jsp and jamon. Needs reconciling, but later.
- Moved DrainingServerTracker and RegionServerTracker to hbase-server:o.a.h.h.master.
- Moved SplitOrMergeTracker to oahh.master (because it depends on a PB)
- Moving hbase-client:oahh.zookeeper.* to hbase-zookeeper module. After HBASE-19200, hbase-client doesn't need them anymore (except 3 classes).
- Renamed some classes to use a consistent naming for classes - ZK instead of mix of ZK, Zk , ZooKeeper. Couldn't rename following public classes: MiniZooKeeperCluster, ZooKeeperConnectionException. Left RecoverableZooKeeper for lack of better name. (suggestions?)
- Sadly, can't move tests out because they depend on HBaseTestingUtility (which defeats part of the purpose - trimming down hbase-server tests. We need to promote more use of mocks in our tests)
Remove a few unused imports.
Remove TestAsyncRegionAdminApi#testOffline, a test for a condition that
no longer exists (no offlining supported in hbase2).
M hbase-server/src/test/java/org/apache/hadoop/hbase/security/access/TestAccessController3.java
Uncomment cleanup called in test teardown.
Restore testRegionCaching and testMulti to working state (required
fixing move procedure and looking for a new exception).
testClusterStatus is broke because multicast is broken.
Updated HTrace version to 4.2
Created TraceUtil class to wrap htrace methods. Uses try with resources.
Signed-off-by: Balazs Meszaros <balazs.meszaros@cloudera.com>
Signed-off-by: Michael Stack <stack@apache.org>
There is no functionality change except for below:
* Variable lastIndexExclusive was getting incremented while locking rows corresponding to input
operations. As a result when getRowLockInternal() method throws TimeoutIOException only operations
in range [nextIndexToProcess, lastIndexExclusive) was getting marked as FAILED before raising
exception up the call stack. With these changes all operations are getting marked as FAILED.
* Cluster Ids of first mutation is used consistently for entire batch. Previous behavior was to use
cluster ids of first mutation in a mini-batch
Signed-off-by: Michael Stack <stack@apache.org>
KeyValue.Type, and its corresponding byte value, are not public API. We
shouldn't have methods that are expecting them. Added a basic sanity
test for isPut and isDelete.
Signed-off-by: Ramkrishna <ramkrishna.s.vasudevan@intel.com>
* pull things that don't rely on HDFS in hbase-server/FSUtils into hbase-common/CommonFSUtils
* refactor setStoragePolicy so that it can move into hbase-common/CommonFSUtils, as a side effect update it for Hadoop 2.8,3.0+
* refactor WALProcedureStore so that it handles its own FS interactions
* add a reflection-based lookup of stream capabilities
* call said lookup in places where we make WALs to make sure hflush/hsync is available.
* javadoc / checkstyle cleanup on changes as flagged by yetus
Signed-off-by: Chia-Ping Tsai <chia7712@gmail.com>
Last mockito-all release was in Dec'14. Mockito-core has had many releases since then.
From mockito's site:
- "Mockito does not produce the mockito-all artifact anymore ; this one was primarily
aimed at ant users, and contained other dependencies. We felt it was time to move on
and remove such artifacts as they cause problems in dependency management system like
maven or gradle."
- anyX() and any(SomeType.class) matchers now reject nulls and check type.
'bypass' logic case by case
Changes Coprocessor ObserverContext 'bypass' semantic. We flip the
default so bypass is NOT supported on Observer invocations; only a
couple of preXXX methods in RegionObserver allow it: e.g. preGet
and prePut but not preFlush, etc. Everywhere else, we throw
a DoesNotSupportBypassException if a Coprocessor Observer
tries to invoke bypass. Master Observers can no longer stop
or change move, split, assign, create table, etc.
Ditto on complete, the mechanism that allowed a Coprocessor
rule that all subsequent Coprocessors are skipped in an
invocation chain; now, complete is only available to
bypassable methods (and Coprocessors will get an exception if
they try to 'complete' when it is not allowed).
See javadoc for whether a Coprocessor Observer method supports
'bypass'. If no mention, 'bypass' is NOT supported.
M hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/CoprocessorHost.java
Added passing of 'bypassable' (and 'completable') and default 'result' argument to
the Operation constructors rather than pass the excecution engine as parameters.
Makes it so can clean up RegionObserverHost and make the calling
clearer regards what is going on.
Methods that support 'bypass' must set this flag on the Observer.
M hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegion.java
Refactoring in here is minor. A few methods that used support bypass
no longer do so removed the check and the need of an if/else meant a
left-shift in some code.
M hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RSRpcServices.java
Ditto
M hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/RegionCoprocessorHost.java
In here label explicitly those methods that are bypassable.
Some changes to make sure we call the corresponding execOperation.
TestMasterObserver had a bunch of test of bypass method. All removed or
disabled.
TODO: What to do w/ the Scanner methods.
Deprecate Table::existsAll method and add Table::exists.
RemoteHTable already had a deprecated exists method, remove that
and implement the new exists from Table interface.
Signed-off-by: Chia-Ping Tsai <chia7712@gmail.com>
Added filterCell method to Filter, it calls filterKeyValue by default
Deprecated filterKeyValue in Filter, bud added default functionality to return Filter.ReturnCode.INCLUDE.
Added filterKeyValue (calling filterCell) to Filters extending FilterBase to be backward compatible.
renamed filterKeyValue to filterCell in all implementations
changed all internal calls to use filterCell instead of filterKeyValue
changed tests too
This way the change is simple and backward compatible.
Any existing custom filter should work since they override filterKeyValue
and the implementation is called by Filter.filterCell.
Moved FilterWrapper to hbase-server
Signed-off-by: anoopsamjohn <anoopsamjohn@gmail.com>
As part of HBASE-18978 the rpc timeout methods gets aligned
between Table and AsyncTable interfaces.
Deprecate the following methods in Table:
- int getRpcTimeout()
- int getReadRpcTimeout()
- int getWriteRpcTimeout()
- int getOperationTimeout()
Add the following methods to Table:
- long getRpcTimeout(TimeUnit)
- long getReadRpcTimeout(TimeUnit)
- long getWriteRpcTimeout(TimeUnit)
- long getOperationTimeout(TimeUnit)
Fix some javadoc issues.
Signed-off-by: Michael Stack <stack@apache.org>
Change name of Interface OnlineRegions to MutableOnlineRegions.
Change name of Interface ImmutableOnlineRegions to OnlineRegions.
Did this since OnlineRegions is for consumer other than internals.
Add a getOnlineRegions to the RegionCoprocessorEnvironment and to
RegionServerCoprocessorEnvironment so CPs can 'access' local
Regions directly.
- Merged BaseCSM class into CSM interface
- Removed config hbase.coordinated.state.manager.class
- Since state manager is not pluggable anymore, we don't need start/stop/initialize to setup unknown classes. Our internal ZkCSM now requires Server in constructor itself. Makes the dependency clearer too.
- Removed CSM from HRegionServer and HMaster constructor. Although it's a step back from dependency injection, but it's more consistent with our current (not good) pattern where we initialize everything in the ctor itself.
Change-Id: Ifca06bb354adec5b11ea1bad4707e014410491fc
Breaks MemStoreSize into MemStoreSize (read-only) and MemStoreSizing
(read/write). MemStoreSize we allow to Coprocesors. MemStoreSizing we
use internally doing MemStore accounting.
Patch to start a standalone RegionServer that register's itself and
optionally stands up Services. Can work w/o a Master in the mix.
Useful testing. Also can be used by hbase-indexer to put up a
Replication sink that extends public-facing APIs w/o need to extend
internals. See JIRA release note for detail.
This patch adds booleans for whether to start Admin and Client Service.
Other refactoring moves all thread and service start into the one fat
location so we can ask to by-pass 'services' if we don't need them.
See JIRA for an example hbase-server.xml that has config to shutdown
WAL, cache, etc.
Adds checks if a service/thread has been setup before going to use it.
Renames the ExecutorService in HRegionServer from service to
executorService.
See JIRA too for example Connection implementation that makes use of
Connection plugin point to receive a replication stream. The default
replication sink catches the incoming replication stream, undoes the
WALEdits and then creates a Table to call a batch with the
edits; up on JIRA, an example Connection plugin (legit, supported)
returns a Table with an overridden batch method where in we do index
inserts returning appropriate results to keep the replication engine
ticking over.
Upsides: an unadulterated RegionServer that will keep replication metrics
and even hosts a web UI if wanted. No hacks. Just ordained configs
shutting down unused services. Injection of the indexing function at a
blessed point with no pollution by hbase internals; only public imports.
No user of Private nor LimitedPrivate classes.
A hack to "hide" the protobufs, but it's not going to be a trivial
change to remove use of protobufs entirely as they're serialized
into the hbase:quota table.
Purges Server, MasterServices, and RegionServerServices from
CoprocessorEnvironments. Replaces removed functionality with
a set of carefully curated methods on the *CoprocessorEnvironment
implementations (Varies by CoprocessorEnvironment in that the
MasterCoprocessorEnvironment has Master-type facility exposed,
and so on).
A few core Coprocessors that should long ago have been converted
to be integral, violate their context; e.g. a RegionCoprocessor
wants free access to a hosting RegionServer (which may or may not
be present). Rather than let these violators make us corrupte the
CP API, instead, we've made up a hacky system that allows core
Coprocessors access to internals. A new CoreCoprocessor Annotation
has been introduced. When loading Coprocessors, if the instance is
annotated CoreCoprocessor, we pass it an Environment that has been
padded w/ extra-stuff. On invocation, CoreCoprocessors know how to
route their way to these extras in their environment.
See the *CoprocessoHost for how the do the check for CoreCoprocessor
and pass a fatter *Coprocessor, one that allows getting of either
a RegionServerService or MasterService out of the environment
via Marker Interfaces.
Removed org.apache.hadoop.hbase.regionserver.CoprocessorRegionServerServices
M hbase-endpoint/src/main/java/org/apache/hadoop/hbase/security/access/SecureBulkLoadEndpoint.java
This Endpoint has been deprecated because its functionality has been
moved to core. Marking it a CoreCoprocessor in the meantime to
minimize change.
M hbase-rsgroup/src/main/java/org/apache/hadoop/hbase/rsgroup/RSGroupAdminEndpoint.java
This should be integral to hbase. Meantime, marking it CoreCoprocessor.
M hbase-server/src/main/java/org/apache/hadoop/hbase/Server.java
Added doc on where it is used and added back a few methods we'd
removed.
A hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/CoreCoprocessor.java
New annotation for core hbase coprocessors. They get richer environment
on coprocessor loading.
A hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/HasMasterServices.java
A hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/HasRegionServerServices.java
Marker Interface to access extras if present.
M hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/MasterCoprocessorEnvironment.java
Purge MasterServices access. Allow CPs a Connection.
M hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/RegionCoprocessorEnvironment.java
Purge RegionServerServices access. Allow CPs a Connection.
M hbase-server/src/main/java/org/apache/hadoop/hbase/coprocessor/RegionServerCoprocessorEnvironment.java
Purge MasterServices access. Allow CPs a Connection.
M hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/MasterSpaceQuotaObserver.java
M hbase-server/src/main/java/org/apache/hadoop/hbase/quotas/QuotaCache.java
We no longer have access to MasterServices. Don't need it actually.
Use short-circuiting Admin instead.
D hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/CoprocessorRegionServerServices.java
Removed. Not needed now we do CP Env differently.
M hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/HRegionServer.java
No need to go via RSS to getOnlineTables; just use HRS.
And so on. Adds tests to ensure we can only get at extra info
if the CP has been properly marked.
* Change imports from org.codehaus to com.fasterxml
* Exclude transitive jackson1 from hadoop and others
* Minor test cleanup to add assert messages, fix some parameter order
* Add anti-pattern check for using jackson 1 imports
* Add explicit non-null serialization directive to ScannerModel
- Table moving to RSG was buggy, because it left the table unassigned.
Now it is fixed we immediately assign to an appropriate RS
(MoveRegionProcedure).
- Table was locked while moving, but unassign operation hung, because
locked table queues are not scheduled while locked. Fixed.
- ProcedureSyncWait was buggy, because it searched the procId in
executor, but executor does not store the return values of internal
operations (they are stored, but immediately removed by the cleaner).
- list_rsgroups in the shell show also the assigned tables and servers.
Signed-off-by: Michael Stack <stack@apache.org>
* batch validation and preparation is done before we start iterating over operations for writes
* durability, familyCellMaps and observedExceptions are batch wide and are now sotred in BatchOperation,
as a result durability is consistent across all operations in a batch
* for all operations done by preBatchMutate() CP hook, operation status is updated to success
* doWALAppend() is modified to habdle replay and is used from doMiniBatchMutate()
* minor improvements
Signed-off-by: Michael Stack <stack@apache.org>
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>
Michael Stack
Apekshit Sharma
huaxiangsun
Zach York
BELUGA BEHR
Chia-Ping Tsai
anoopsamjohn
Andrew Purtell
TAK LON WU
tedyu
zhangduo
Guanghao Zhang
Josh Elser
Guangxu Cheng
Harshal Deepakkumar Jain
Toshihiro Suzuki
binlijin
Jerry He
Balazs Meszaros
Peter Somogyi
Vasudevan
Reid Chan
Gary Helmling
Apekshit
huzheng
Umesh Agashe
Sean Busbey
Lars Hofhansl
Yi Liang
Mike Drob
Tamas Penzes
QilinCao
Michael Stack
churro
eshcar
anastas
Sreeram Venkatasubramanian
xiaowen147
Xiang Li
Jan Hentschel
Maytee Chinavanichkit
libisthanks
libisthanks
Ashu Pachauri
Andy Yang
Pankaj Kumar
Chinmay Kulkarni
Abhishek Singh Chouhan