* HBASE-22749 Distributed MOB compactions
- 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.
* HBASE-22749 Distributed MOB compactions
fix RestrictedApi
Co-authored-by: Vladimir Rodionov <vrodionov@apache.org>
Signed-off-by: Wellington Chevreuil <wchevreuil@apache.org>
TestMultiRespectLimits submits a put with an empty value. This appears
to not be what was intended. Instead, use a single byte.
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Replace BufferChain#write(channel,int) with a simpler #write(channel)
implementation that does not attempt to "chunk" data to be written. This
method was used exclusively by SimpleRpcServer. The code was unnecessarily
complex and caused short writes when values were large, so was corrected
and simplified. Any difference in performance from this change will be
limited to SimpleRpcServer. Testing under load confirms the fix and does
not show significant regression.
SimpleRpcServer and its related code is now also marked as @Deprecated.
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Conflicts:
hbase-server/src/main/java/org/apache/hadoop/hbase/ipc/SimpleRpcServer.java
Co-authored-by: Ishika Soni <isoni@isoni-ltmag9r.internal.salesforce.com>
Signed-off-by: Duo Zhang <zhangduo@apache.org>
(cherry picked from commit acfbc3ff7d)
A larger IO buffer for absorbing WALCodec writes can improve the compression
ratio of larger values, because the compressor will be given a larger internal
buffer over which there will be more match opportunities. Does not impact the
ability to read existing written files.
Also, reset the BAOS internal buffer on the way out of compress() so potential
large-ish buffers do not linger on the heap longer than necessary.
Signed-off-by: Duo Zhang <zhangduo@apache.org>
Support site configuration of the bytebuf allocator that Netty will use for
NettyRpcServer channels. Property name is 'hbase.netty.rpcserver.allocator'.
Default is no value, which is equivalent to "pooled". Valid values are:
- "pooled": use PooledByteBufAllocator
- "unpooled": use UnpooledByteBufAllocator
- "heap": use HeapByteBufAllocator, which is a PooledByteBufAllocator that
preferentially allocates buffers on heap wherever possible
- <class>: If the value is none of the recognized labels, treat it as a class
name implementing org.apache.hbase.thirdparty.io.netty.buffer.ByteBufAllocator.
This allows the user to add a custom implementation, perhaps for debugging.
Also updates ReflectionUtils with a new helper method.
Signed-off-by: Viraj Jasani <vjasani@apache.org>
Conflicts:
hbase-server/src/main/java/org/apache/hadoop/hbase/ipc/NettyRpcServer.java
hbase-server/src/test/java/org/apache/hadoop/hbase/ipc/TestNettyRpcServer.java
This reverts commit f3a48d1910.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
Conflicts:
hbase-server/src/test/java/org/apache/hadoop/hbase/regionserver/TestHRegion.java
In #4470 for HBASE-26192, it was noted that the HbckChore is kind of a pain to use and test
because it maintains a bunch of local state. By contract, the CatalogJanitorChore makes a nice
self-contained report. Let's update HbckChore to do the same.
Signed-off-by: Andrew Purtell <apurtell@apache.org>
BukrosSzabolcs
Duo Zhang
Andrew Purtell
Bryan Beaudreault
chenglei
Ujjawal
tianhang
Ishika
Kerasone
Pankaj
litao
LiangJun He
xiaozhang0319
SiCheng-Zheng
Xiaolin Ha
Nick Dimiduk
Viraj Jasani
liuzhuang2017
huaxiangsun
skysiders