History log of /6.0.3/forestdb/ (Results 151 - 175 of 1302)
Revision (<<< Hide revision tags) (Show revision tags >>>)Date Author Comments
(<<< Hide modified files)
(Show modified files >>>)
83eb6aa823-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Disable compaction and rollback interleaving test

There is a sporadic failure in testing interleaving compaction and rollback
task. Disabling it until we resolve this test issue.

Disable compaction and rollback interleaving test

There is a sporadic failure in testing interleaving compaction and rollback
task. Disabling it until we resolve this test issue.

Change-Id: I2cae581cd91ed2778cfc40d10be7aea6c8e3b795
Reviewed-on: http://review.couchbase.org/61901
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

86594dc023-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18747: Temporarily disable WAL sharing

WAL sharing introduces slightly increased latencies in
query path causing crashes in gsi. Temporarily disabling
to unblock testing until it

MB-18747: Temporarily disable WAL sharing

WAL sharing introduces slightly increased latencies in
query path causing crashes in gsi. Temporarily disabling
to unblock testing until it is resolved.
compile with -D_MVCC_WAL=1 to re-enable.

Change-Id: I40e461202009db124c6d4b29a905bebc3e235007
Reviewed-on: http://review.couchbase.org/61345
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

532e494823-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18747 Don't grab the spinlock when incrementing a file ref counter

As a file ref counter is an atomic variable, we don't need to grab the file
manager's spinlock when the file is alre

MB-18747 Don't grab the spinlock when incrementing a file ref counter

As a file ref counter is an atomic variable, we don't need to grab the file
manager's spinlock when the file is already opened and its ref counter only
needs to be incremented.

Change-Id: Ie49b0202e8771b1ab57105d864cadf9cc1d23ee8
Reviewed-on: http://review.couchbase.org/61894
Reviewed-by: Sundararaman Sridharan <sundar@couchbase.com>
Tested-by: buildbot <build@couchbase.com>

show more ...

b57342cc23-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Fix compact_functional_test failure

Change-Id: I0cf98e2ca01fad303d44613336aab5742b2cdfe2
Reviewed-on: http://review.couchbase.org/61866
Tested-by: buildbot <build@couchbase.com>

Fix compact_functional_test failure

Change-Id: I0cf98e2ca01fad303d44613336aab5742b2cdfe2
Reviewed-on: http://review.couchbase.org/61866
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

c3eb7ba523-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18747 Don't grab spinlock when decrementing a file's ref counter

As a file ref counter is an atomic variable, we don't need to grab
the spinlock upon decrementing it.

Change-

MB-18747 Don't grab spinlock when decrementing a file's ref counter

As a file ref counter is an atomic variable, we don't need to grab
the spinlock upon decrementing it.

Change-Id: I07bd6010bb5df9f34ce9417e78f12273742c11d5
Reviewed-on: http://review.couchbase.org/61862
Reviewed-by: Jung-Sang Ahn <jungsang.ahn@gmail.com>
Tested-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

923d253c23-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

MB-18846: Update minimum cmake version required

Minumum CMake version required to support OBJECT
library types is 2.8.8.

Change-Id: Ic4dae5d41b6c4790eba2fec89cb090d0907f46cf

MB-18846: Update minimum cmake version required

Minumum CMake version required to support OBJECT
library types is 2.8.8.

Change-Id: Ic4dae5d41b6c4790eba2fec89cb090d0907f46cf
Reviewed-on: http://review.couchbase.org/61861
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
Tested-by: buildbot <build@couchbase.com>

show more ...

1abf325c22-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

Real Use Case Simulation with Configurable Disk behaviour

This adds a disk_sim_test to simulate real use cases such
as GSI but with an unpredictable or anomalous disk.
Real use case

Real Use Case Simulation with Configurable Disk behaviour

This adds a disk_sim_test to simulate real use cases such
as GSI but with an unpredictable or anomalous disk.
Real use case also carries accurate validation of database.

Current tests:
+indexer Stale=ok pattern small batch size and
frequent snapshots
- 1 writer 7 iterators, all clone snapshots
handles

Change-Id: If12cda570842655fcf2f240b424483d6b2454508
Reviewed-on: http://review.couchbase.org/61543
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

2641acc522-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18753: Retain offset of immediately deleted docs

During wal restore, transactional docs are restored based
on the offset in the commit mark doc. For deleted and immediately
purged

MB-18753: Retain offset of immediately deleted docs

During wal restore, transactional docs are restored based
on the offset in the commit mark doc. For deleted and immediately
purged docs, we need this value to be correct otherwise deleted
and immediately purged transactional docs will be lost on reopen.

Change-Id: Ia01e3fbec1106eb01a896d7709fbd410eaaee708
Reviewed-on: http://review.couchbase.org/61788
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>
Tested-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

f7c15a5c22-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18747 Get rid of locking on closing dirty index node snapshot

- Now read-only thread will not grab lock during _fdb_kvs_close().

- And also minimized any critical sections relate

MB-18747 Get rid of locking on closing dirty index node snapshot

- Now read-only thread will not grab lock during _fdb_kvs_close().

- And also minimized any critical sections related to handling dirty
index node snapshots.

Change-Id: I33e0c95b5b739fdf82d8835c784eebf75067dd38
Reviewed-on: http://review.couchbase.org/61807
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

bf67fcb222-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18765 Abort the 2nd phase of compaction on rollback request

- Added the missing abort routine for rollback/compaction cancel
in the second phase of compaction (moving delta)

MB-18765 Abort the 2nd phase of compaction on rollback request

- Added the missing abort routine for rollback/compaction cancel
in the second phase of compaction (moving delta)

Change-Id: I25ad9fff8524ce279f195217fd2b1e56173bef9b
Reviewed-on: http://review.couchbase.org/61799
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

889f945d22-Mar-2016 Trond Norbye <trond.norbye@gmail.com>

alloca needs include of alloca.h on Solaris

Change-Id: I0bd555be99d3de56400b8edcb56c41e2d0095f50
Reviewed-on: http://review.couchbase.org/61808
Reviewed-by: Sergey Avseyev <sergey.av

alloca needs include of alloca.h on Solaris

Change-Id: I0bd555be99d3de56400b8edcb56c41e2d0095f50
Reviewed-on: http://review.couchbase.org/61808
Reviewed-by: Sergey Avseyev <sergey.avseyev@gmail.com>
Tested-by: buildbot <build@couchbase.com>

show more ...

8d6ce93519-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

Adding fdbench-micro to forestdb/tests

Integrating micro benchmarks for forestdb into
the main repository. These will be run as part
of "make test" during CV.

Sample output

Adding fdbench-micro to forestdb/tests

Integrating micro benchmarks for forestdb into
the main repository. These will be run as part
of "make test" during CV.

Sample output for iterator test latencies:

========== Avg Latencies (ITERATOR_TEST_STATS) - 256 samples (µs) ====================

Percentile
Median 95th 99th Std Dev Histogram of samples
iterator_init 22.000 38.000 163.000 22.260 ▁▁▁▁▁▁▁▁▁▂▂▃▃▅▆▄▄▅▄▃▃▃▃▅▇█▅▄▃▃▄▄
iterator_next 1.000 2.000 3.000 0.737 ▁▃█▃▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
iterator_get 3.000 5.000 6.000 1.316 ▁▁▁▂█▁▅▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
iterator_close 2.000 3.000 5.000 2.032 ▁▂█▇▄▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁▁
0 µs 38
=======================================================================================

Change-Id: I43bf7830585837fec5628628063d06dc6bf7bdf6
Reviewed-on: http://review.couchbase.org/61466
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

25638c9e17-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

Adding class StatsAggregator to tests

This functionality is for collecting and aggregating
stats. Aggregated stats' median, percentiles and
histograms can be generated as well.

Adding class StatsAggregator to tests

This functionality is for collecting and aggregating
stats. Aggregated stats' median, percentiles and
histograms can be generated as well.

Change-Id: I55f11b2d3a29291da6c52c15df6bc6b3765f917b
Reviewed-on: http://review.couchbase.org/61725
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

f4ad491f21-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18594 Avoid duplicate items during WAL flushing

- get_old_offset() may return same 'old_offset' value on different
keys; this is because hbtrie_find_offset() (which is internally

MB-18594 Avoid duplicate items during WAL flushing

- get_old_offset() may return same 'old_offset' value on different
keys; this is because hbtrie_find_offset() (which is internally
called by get_old_offset()) does not compare entire key string but
just prefix only, due to performance issue.

- As a result, aforementioned case (keys are different but both
'old_offset' and 'offset' are same) very rarely happens so that
different items are duplicate, and causes crash.

- To avoid this issue, we need to additionally compare sequence
numbers so as to distinguish those two different items.

Change-Id: I697f12422c70c60d292cbfe765f9e06b7c02b0e2
Reviewed-on: http://review.couchbase.org/61771
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

688923d319-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18747: Don't lock on wal snapshot close

GSI tests like stale=ok are extremely sensitive to higher
latencies in the query path. If during wal_flush the snapshots
are closed in a hi

MB-18747: Don't lock on wal snapshot close

GSI tests like stale=ok are extremely sensitive to higher
latencies in the query path. If during wal_flush the snapshots
are closed in a highly concurrent query thread, the query
thread stalls, causing go to spawn more threads and eventually
running out of threads. Simple fix is to remove the unnecessary
locking in the close path.

Change-Id: Ica601d3420eb3be099f26f0ff141967518975859
Reviewed-on: http://review.couchbase.org/61747
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

bc18d4ed18-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18747 Don't grab the spinlock in incrementing a ref counter

A reference counter for filemgr or filemgr_dirty_update_node can be incremented
without grabbing a spinlock because its ref

MB-18747 Don't grab the spinlock in incrementing a ref counter

A reference counter for filemgr or filemgr_dirty_update_node can be incremented
without grabbing a spinlock because its reference counter is already greater than
zero due to the existence of the source filemgr or filemgr_dirty_update_node.
Therefore, an atomic type should be enough to maintain their reference counters.

Change-Id: I130991a41e5ddc7ec4fd1b30f04e2be65c7dc367
Reviewed-on: http://review.couchbase.org/61691
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

aa8ae3ab17-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17904 Make C++11 support as a required build environment

As we plan to move our code repository to C++11 incrementally, this
commit makes C++11 support as a build environment required

MB-17904 Make C++11 support as a required build environment

As we plan to move our code repository to C++11 incrementally, this
commit makes C++11 support as a build environment required.

Change-Id: Iba2f8ceed33af635e61647b6522169c74e7cfed1
Reviewed-on: http://review.couchbase.org/61663
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

b165bcd617-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17904 Use relaxed_memory_order if a strict consistent order is not required

If a given atomic variable doesn't require a strict consistent memory order,
then relaxed memory order shou

MB-17904 Use relaxed_memory_order if a strict consistent order is not required

If a given atomic variable doesn't require a strict consistent memory order,
then relaxed memory order should be enough for atomic operations. For example,
operation stats and some config parameters can use relaxed memory order.

Change-Id: Ic1454c5d23da64fbe5309968026938bdd6803d27
Reviewed-on: http://review.couchbase.org/61624
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

c83d983316-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17904 Use C++11 atomic library if supported

If a build environment has C++11 support, then this change allows
us to use C++11 atomic library instead of other built-in implementations.

MB-17904 Use C++11 atomic library if supported

If a build environment has C++11 support, then this change allows
us to use C++11 atomic library instead of other built-in implementations.

The subsequent change will provide an atomic operation with a relaxed
memory order from C++11.

Change-Id: I9f26ceddb42c68f00deecf5518bfba388ff4c217
Reviewed-on: http://review.couchbase.org/61605
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

Revision tags: v1.1
24b5f63814-Mar-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Deprecate atomic_destroy functions as they are no longer needed

As atomic_init API doesn't allocate any resources, we don't need
atomic_destroy functions anymore.

Change-Id: I45

Deprecate atomic_destroy functions as they are no longer needed

As atomic_init API doesn't allocate any resources, we don't need
atomic_destroy functions anymore.

Change-Id: I45f27d91bd4c5473166cb61c40cda73e6aaec0cf
Reviewed-on: http://review.couchbase.org/61604
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

daecda9516-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

Use $<TARGET_OBJECTS:..> in tests/../CMakeLists for uniformity

For uniformity, instead of FORESTDB_CORE_SRC and
FORESTDB_UTILS_SRC use $<TARGET_OBJECTS:FDB_TOOLS_CORE>
and $<TARGET_O

Use $<TARGET_OBJECTS:..> in tests/../CMakeLists for uniformity

For uniformity, instead of FORESTDB_CORE_SRC and
FORESTDB_UTILS_SRC use $<TARGET_OBJECTS:FDB_TOOLS_CORE>
and $<TARGET_OBJECTS:FDB_TOOLS_UTILS>.

Change-Id: I7cb4826f0a0e028e79fb040f6beac8f05922be51
Reviewed-on: http://review.couchbase.org/61589
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

22c2d01c15-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

Enabling code_coverage for forestDb

A new Jenkins nightly job will now generate code
coverage results for forestdb, and this change is
required for it to run successfully.

H

Enabling code_coverage for forestDb

A new Jenkins nightly job will now generate code
coverage results for forestdb, and this change is
required for it to run successfully.

Here's the link to the jenkins project:
http://factory.couchbase.com/view/build/view/Code%20coverage/job/forestdb-nightly-coverage/

Change-Id: I1c41724a5a2c63f78e5c9ae65a9871409c9f8fe8
Reviewed-on: http://review.couchbase.org/61521
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

c1932afe15-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18712 Fix to a bug on comparison between NULL and zero-filled chunk

- In the current HB+trie, it cannot distinguish NULL and zero-filled
chunks because they are same (0x0000...) in th

MB-18712 Fix to a bug on comparison between NULL and zero-filled chunk

- In the current HB+trie, it cannot distinguish NULL and zero-filled
chunks because they are same (0x0000...) in the disk node.

- It causes an issue when
1) an iterator is initialized with a sub-string of the smallest key, and
2) the initial key is exactly aligned to chunk size, and
3) the rest of string excluding the prefix are all zero.

- To address this issue, we need to handle that case separately during
iterator initialization.

Change-Id: Ib85e7fa232860af058924095f7e5fee3cddeff33
Reviewed-on: http://review.couchbase.org/61505
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

4b59f4f114-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

[2] Adding End-to-end tests for stale block reuse

tests for:
+ crash testing
+ multi reader writers
+ multi dbfile

Change-Id: Ibeeffbc418bede3e922e9107cb08a4eeb38273df

[2] Adding End-to-end tests for stale block reuse

tests for:
+ crash testing
+ multi reader writers
+ multi dbfile

Change-Id: Ibeeffbc418bede3e922e9107cb08a4eeb38273df
Reviewed-on: http://review.couchbase.org/58829
Tested-by: buildbot <build@couchbase.com>
Reviewed-by: Chiyoung Seo <chiyoung@couchbase.com>

show more ...

8312f1eb14-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

Revert "Do not cache DB header in blockcache for performance"

This reverts commit c29d383cbd6f967c18e6fbffb4b83d69cd1bf65f.

12345678910>>...53