Depending on which compression codec is used, a short read of the
compressed bytes can cause catastrophic errors that confuse the WAL reader.
This problem can manifest when the reader is actively tailing the WAL for
replication. To avoid these issues when WAL value compression is enabled,
BoundedDelegatingInputStream should assume enough bytes are available to
supply a reader up to its bound. This behavior is valid per the contract
of available(), which provides an _estimate_ of available bytes, and
equivalent to IOUtils.readFully but without requiring an intermediate
buffer.
Added TestReplicationCompressedWAL and TestReplicationValueCompressedWAL.
Without the WALCellCodec change TestReplicationValueCompressedWAL will
fail.
Signed-off-by: Bharath Vissapragada <bharathv@apache.org>
We introduced EnvironmentEdgeManager as a way to inject alternate clocks
for unit tests. In order for this to be effective, all callers that would
otherwise use System.currentTimeMillis() must call
EnvironmentEdgeManager.currentTime() instead, except the implementers of
EnvironmentEdge.
Signed-off-by: Bharath Vissapragada <bharathv@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Undo asserts that LZ4 and SNAPPY fails if their native libs are NOT
loaded; as of hadoop 3.3.1, LZ4 and SNAPPY can work w/o native libs.
Signed-off-by: Duo Zhang <zhangduo@apache.org>
WAL storage can be expensive, especially if the cell values
represented in the edits are large, consisting of blobs or
significant lengths of text. Such WALs might need to be kept around
for a fairly long time to satisfy replication constraints on a space
limited (or space-contended) filesystem.
We have a custom dictionary compression scheme for cell metadata that
is engaged when WAL compression is enabled in site configuration.
This is fine for that application, where we can expect the universe
of values and their lengths in the custom dictionaries to be
constrained. For arbitrary cell values it is better to use one of the
available compression codecs, which are suitable for arbitrary albeit
compressible data.
Signed-off-by: Bharath Vissapragada <bharathv@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Signed-off-by: Nick Dimiduk <ndimiduk@apache.org>
Remove the deprecated fields, which can be removed in 3.0.0. Marked the
constant OLDEST_TIMESTAMP as InterfaceAudience.Private as it is only use
in classes, which are also marked as InterfaceAudience.Private.
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
This revert of a revert reapplies the PR; the original application was
missing the HBASE JIRA #; thats why it was reverted and then reapplied
w/ the JIRA # added.
This reverts commit 16fe1e95ec.
Adds "hbase.master.executor.merge.dispatch.threads" and defaults to 2.
Also adds additional logging that includes the number of split plans
and merge plans computed for each normalizer run.
Signed-off-by: Wellington Chevreuil <wchevreuil@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Revert of the revert -- re-applying HBASE-25449 with a change
of renaming the test hdfs XML configuration file as it was adversely
affecting tests using MiniDFS
This reverts commit c218e576fe.
Co-authored-by: Josh Elser <elserj@apache.org>
Signed-off-by: Peter Somogyi <psomogyi@apache.org>
Signed-off-by: Michael Stack <stack@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
* HBASE-25379 Make retry pause time configurable for regionserver short operation RPC (reportRegionStateTransition/reportProcedureDone)
* HBASE-25379 RemoteProcedureResultReporter also should retry after the configured pause time
* Addressed the review comments
Signed-off-by: Yulin Niu <niuyulin@apache.org>
* Using ContiguousCellFormat as a marker alone
* Commit the new file
* Fix the comparator logic that was an oversight
* Fix the sequenceId check order
* Adding few more static methods that helps in scan flow like query
matcher where we have more cols
* Remove ContiguousCellFormat and ensure compare() can be inlined
* applying negation as per review comment
* Fix checkstyle comments
* fix review comments
* Address review comments
* Fix the checkstyle issues
* Fix javadoc
Signed-off-by: stack <stack@apache.org>
Signed-off-by: AnoopSamJohn <anoopsamjohn@apache.org>
Signed-off-by: huaxiangsun <huaxiangsun@apache.org>
Network identities should be bound late. Remote addresses should be
resolved at the last possible moment, just before connect(). Network
identity mappings can change, so our code should not inappropriately
cache them. Otherwise we might miss a change and fail to operate normally.
Revert "HBASE-14544 Allow HConnectionImpl to not refresh the dns on errors"
Removes hbase.resolve.hostnames.on.failure and related code. We always
resolve hostnames, as late as possible.
Preserve InetSocketAddress caching per RPC connection. Avoids potential
lookups per Call.
Replace InetSocketAddress with Address where used as a map key. If we want
to key by hostname and/or resolved address we should be explicit about it.
Using Address chooses mapping by hostname and port only.
Add metrics for potential nameservice resolution attempts, whenever an
InetSocketAddress is instantiated for connect; and metrics for failed
resolution, whenever InetSocketAddress#isUnresolved on the new instance
is true.
* Use ServerName directly to build a stub key
* Resolve and cache ISA on a RpcChannel as late as possible, at first call
* Remove now invalid unit test TestCIBadHostname
We resolve DNS at the latest possible time, at first call, and do not
resolve hostnames for creating stubs at all, so this unit test cannot
work now.
Reviewed-by: Mingliang Liu <liuml07@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
This PR is a follow-up of HBASE-25181 (#2539), where several issues were
discussed on the PR:
1. Currently we use PBKDF2WithHmacSHA1 key generation algorithm to generate a
secret key for HFile / WalFile encryption, when the user is defining a string
encryption key in the hbase shell. This algorithm is not secure enough and
not allowed in certain environments (e.g. on FIPS compliant clusters). We are
changing it to PBKDF2WithHmacSHA384. It will not break backward-compatibility,
as even the tables created by the shell using the new algorithm will be able
to load (e.g. during bulkload / replication) the HFiles serialized with the
key generated by an old algorithm, as the HFiles themselves already contain
the key necessary for their decryption.
Smaller issues fixed by this commit:
2. Improve the documentation e.g. with the changes introduced by HBASE-25181
and also by some points discussed on the Jira ticket of HBASE-25263.
3. In EncryptionUtil.createEncryptionContext the various encryption config
checks should throw IllegalStateExceptions instead of RuntimeExceptions.
4. Test cases in TestEncryptionTest.java should be broken down into smaller
tests.
5. TestEncryptionDisabled.java should use ExpectedException JUnit rule to
validate exceptions.
closes#2676
Signed-off-by: Peter Somogyi <psomogyi@apache.org>
* HBASE-25050 - We initialize Filesystems more than once.
* Ensuring that calling the FS#get() will only ensure FS init.
* Fix for testfailures. We should pass the entire path and no the scheme
alone
* Cases where we don't have a scheme for the URI
* Address review comments
* Add some comments on why FS#get(URI, conf) is getting used
* Adding the comment as per Sean's review
Signed-off-by: Sean Busbey <busbey@apache.org>
Signed-off-by: Michael Stack <stack@apache.org>
* HBASE-25187 Improve SizeCachedKV variants initialization
* HBASE-25187 Improve SizeCachedKV variants initialization
* The BBKeyValue also can be optimized
* Change for SizeCachedKeyValue
* Addressing revew comments
* Fixing checkstyle and spot bugs comments
* Spot bug fix for hashCode
* Minor updates make the rowLen as short and some consturctor formatting
* Change two more places where there was a cast
Prior to this patch hbase always used the MD5 hash algorithm to store a hash for encryption keys.
This hash is needed to verify the secret key of the subject. (e.g. making
sure that the same secrey key is used during encrypted HFile read and write).
The MD5 algorithm is considered weak, and can not be used in some
(e.g. FIPS compliant) clusters.
In this patch we:
- add a config parameter to globally enable/disable column family encryption (def enabled)
- introduce a backward compatible way of specifying the hash algorithm.
This enable us to use newer and more secure hash algorithms like SHA-384
or SHA-512 (which are FIPS compliant).
- add a config parameter to fail if an hfile is encountered that uses a
different hash algorithm than the one currently configured to ease validation after
migrating key hash algorithms (def disabled)
Closes#2539
Signed-off-by: Sean Busbey <busbey@apache.org>
Signed-off-by: Esteban Gutierrez <esteban@apache.org>
Wire up the `ConfigurationObserver` chain for
`RegionNormalizerManager`. The following configuration keys support
hot-reloading:
* hbase.normalizer.throughput.max_bytes_per_sec
* hbase.normalizer.split.enabled
* hbase.normalizer.merge.enabled
* hbase.normalizer.min.region.count
* hbase.normalizer.merge.min_region_age.days
* hbase.normalizer.merge.min_region_size.mb
Note that support for `hbase.normalizer.period` is not provided
here. Support would need to be implemented generally for the `Chore`
subsystem.
Signed-off-by: Bharath Vissapragada <bharathv@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Signed-off-by: Aman Poonia <aman.poonia.29@gmail.com>
The core change here is to the loop in
`SimpleRegionNormalizer#computeMergeNormalizationPlans`. It's a nested
loop that walks the table's region chain once, looking for contiguous
sequences of regions that meet the criteria for merge. The outer loop
tracks the starting point of the next sequence, the inner loop looks
for the end of that sequence. A single sequence becomes an instance of
`MergeNormalizationPlan`.
Signed-off-by: Huaxiang Sun <huaxiangsun@apache.org>
Add the documentation when
HConstants#REPLICATION_DROP_ON_DELETED_TABLE_KEY was deprecated and when
it is expected to be removed.
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Make it so WALPlayer can replay recovered.edits files.
hbase-mapreduce/src/main/java/org/apache/hadoop/hbase/mapreduce/WALInputFormat.java
Allow for WAL files that do NOT have a startime in their name.
Use the 'generic' WAL-filename parser instead of the one that
used be local here. Implement support for 'startTime' filter.
Previous was just not implemented.
hbase-mapreduce/src/main/java/org/apache/hadoop/hbase/mapreduce/WALPlayer.java
Checkstyle.
hbase-server/src/main/java/org/apache/hadoop/hbase/wal/AbstractFSWALProvider.java
Use the new general WAL name timestamp parser.
hbase-server/src/main/java/org/apache/hadoop/hbase/wal/WAL.java
Utility for parsing timestamp from WAL filename.
hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestRecoveredEdits.java
Export attributes about the local recovered.edits test file
so other tests can play with it.
Signed-off-by: Wellington Chevreuil <wchevreuil@apache.org>
* Admin API getLogEntries() for ring buffer use-cases: so far, provides balancerDecision and slowLogResponse
* Refactor RPC call for similar use-cases
* Single RPC API getLogEntries() for both Master.proto and Admin.proto
Closes#2261
Signed-off-by: Andrew Purtell <apurtell@apache.org>
* Break subclass referencing of MetaCellComparator from superclass CellComparatorImpl
static initializer by moving META_COMPARATOR to subclass MetaCellComparator
Closes#2329
Signed-off-by: Duo Zhang <zhangduo@apache.org>
when neighbor is larger than average size
* add `testMergeEmptyRegions` to explicitly cover different
interleaving of 0-sized regions.
* fix bug where merging a 0-size region is skipped due to large
neighbor.
* remove unused `splitPoint` from `SplitNormalizationPlan`.
* generate `toString`, `hashCode`, and `equals` methods from Apache
Commons Lang3 template on `SplitNormalizationPlan` and
`MergeNormalizationPlan`.
* simplify test to use equality matching over `*NormalizationPlan`
instances as plain pojos.
* test make use of this handy `TableNameTestRule`.
* fix line-length issues in `TestSimpleRegionNormalizer`
Signed-off-by: Wellington Chevreuil <wchevreuil@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Signed-off-by: huaxiangsun <huaxiangsun@apache.org>
Signed-off-by: Aman Poonia <aman.poonia.29@gmail.com>
* refactor how we use connection and async connection to rely on their access methods
* refactor initialization and cleanup of the shared connection
* incompatibly change HCTU's Configuration member variable to be final so it can be safely accessed from multiple threads.
Closes#2180
Signed-off-by: Wellington Chevreuil <wchevreuil@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Also fix three bugs:
* We were trying to delete non-empty directory; weren't doing
accounting for meta WALs where meta had moved off the server
(successfully)
* We were deleting split WALs rather than archiving them.
* We were not handling corrupt files.
Deprecations and removal of tests of old system.
Signed-off-by: Anoop Sam John <anoopsamjohn@apache.org>
Writing a test for this is tricky. There is enough coverage for
functional tests. Only concern is performance, but there is enough
logging for it to detect timed out/badly performing sync calls.
Additionally, this patch decouples the ZK event processing into it's
own thread rather than doing it in the EventThread's context. That
avoids deadlocks and stalls of the event thread.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Simplify our Normalizer story to have just a single, configurable
implementation.
* fold the features of `MergeNormalizer` into
`SimpleRegionNormalizer`, removing the intermediate abstract class.
* configuration keys for merge-only features now share a common
structure.
* add configuration to selectively disable normalizer split/merge
operations.
* `RegionNormalizer` now extends `Configurable` instead of creating a
new instance of `HBaseConfiguration` or snooping one off of other
fields.
* avoid the extra RPCs by using `MasterServices` instead of
`MasterRpcServices`.
* boost test coverage of all the various flags and feature
combinations.
Signed-off-by: Michael Stack <stack@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Signed-off-by: huaxiangsun <huaxiangsun@apache.org>
This utility is useful for any module that wants to detect
dynamic config changes. Having it to hbase-common makes it
accessible to all the other modules.
Signed-off-by: Michael Stack <stack@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
This is needed for the hbase-native-client to compile. Also ships
the includes in the assembly tar.
Signed-off-by: Marc <phrocker@apache.org>
Signed-off-by: Michael Stack <stack@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Signed-off-by: Anoop Sam John <anoopsamjohn@apache.org>
Signed-off-by: ramkrish86 <ramkrishna@apache.org>
Simplify the new user experience shipping a configuration that enables
a fresh checkout or tarball distribution to run in standalone mode
without direct user configuration. This change restores the behavior
we had when running on Hadoop 2.8 and earlier.
Patch for master includes an update to the book. This change will be
omitted when backporting to earlier branches.
Signed-off-by: stack <stack@apache.org>
Signed-off-by: Josh Elser <elserj@apache.org>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Move the random free port generate back into hbasecommontestingutility
from hbasetestingutility.
Add a create simple kdc server utility that will start a kdc server and
if a bindexception, create a new one on a new random port in hbase-common.
Add new BoundSocketMaker helpful when trying to manufacture
BindExceptions because of port clash.
Change thrift and http kdc tests to use this new utility (removes
code duplication around kdc server setup).
* Reorganize MOB compaction tests for more reuse.
* Add tests for mob compaction after snapshot clone operations
* note the original table used to write a given mob hfile and use that to find it later.
Signed-off-by: Esteban Gutierrez <esteban@apache.org>
Master rpc server end point doesn't bind to localhost's
IP address by default. Instead, it looks up the hostname and
binds to the endpoint to which it resolves. MasterRegistry should
do the same when building the default server end point to talk to.
Signed-off-by: Nick Dimiduk <ndimiduk@apache.org>
There were a couple of issues.
- There was a leak of a file descriptor for hbck lock file. This
was contributing to all the "ConnectionRefused" stack traces since
it was trying to renew lease for an already expired mini dfs cluster.
This issue was there for a while, just that we noticed it now.
- After upgrade to JUnit 4.13, it looks like the behavior for test
timeouts has changed. Earlier the timeout seems to have applied for
each parameterized run, but now it looks like it is applied across
all the runs.
This patch fixes both the issues.
Signed-off-by: Stack <stack@apache.org>
Signed-off-by: Jan Hentschel <jan.hentschel@ultratendency.com>
Implements a master based registry for clients.
- Supports hedged RPCs (fan out configured via configs).
- Parameterized existing client tests to run with multiple registry combinations.
- Added unit-test coverage for the new registry implementation.
Signed-off-by: Nick Dimiduk <ndimiduk@apache.org>
Signed-off-by: stack <stack@apache.org>
Signed-off-by: Andrew Purtell <apurtell@apache.org>
- MOB compaction is now handled in-line with per-region compaction on region
servers
- regions with mob data store per-hfile metadata about which mob hfiles are
referenced
- admin requested major compaction will also rewrite MOB files; periodic RS
initiated major compaction will not
- periodically a chore in the master will initiate a major compaction that
will rewrite MOB values to ensure it happens. controlled by
'hbase.mob.compaction.chore.period'. default is weekly
- control how many RS the chore requests major compaction on in parallel
with 'hbase.mob.major.compaction.region.batch.size'. default is as
parallel as possible.
- periodic chore in master will scan backing hfiles from regions to get the
set of referenced mob hfiles and archive those that are no longer
referenced. control period with 'hbase.master.mob.cleaner.period'
- Optionally, RS that are compacting mob files can limit write
amplification by not rewriting values from mob hfiles over a certain size
limit. opt-in by setting 'hbase.mob.compaction.type' to 'optimized'.
control threshold by 'hbase.mob.compactions.max.file.size'.
default is 1GiB
- Should smoothly integrate with existing MOB users via rolling upgrade.
will delay old MOB file cleanup until per-region compaction has managed
to compact each region at least once so that used mob hfile metadata can
be gathered.
meiyi
Duo Zhang
Aman Poonia
Xiaolin Ha
Wei-Chiu Chuang
Peter Somogyi
Andrew Purtell
Michael Stack
Baiqiang Zhao
GeorryHuang
ZhiChen
Jan Hentschel
haxiaolin
bitterfox
shahrs87
niuyulin
stack
zhengzhuobinzzb
Bharath Vissapragada
申胜利
Pankaj
ramkrish86
Adam
Mate Szalay-Beko
gvprathyusha6
WenFeiYi
Nick Dimiduk
sanjeetnishad95
niuyulin
Guanghao Zhang
Viraj Jasani
XinSun
bsglz
Sean Busbey
Bo Cui
pengmq1
Anoop Sam John
Mingliang Liu
Sambit Mohapatra
Yu Wang
Mark Robert Miller
Reid Chan
Abhinaba Sarkar
Vladimir Rodionov