History log of /6.0.3/forestdb/ (Results 201 - 225 of 1302)
Revision (<<< Hide revision tags) (Show revision tags >>>)Date Author Comments
(<<< Hide modified files)
(Show modified files >>>)
06fb3e3a03-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18364: WAL flush one item/key & retain deletes till root update

With MVCC WAL, a key can have multiple items, WAL flush must
only pick one unique item per key, else there can be doubl

MB-18364: WAL flush one item/key & retain deletes till root update

With MVCC WAL, a key can have multiple items, WAL flush must
only pick one unique item per key, else there can be double free crash.
If purge_interval=0, deleted items in WAL which are not in
main index were dropped immediately. If a snapshot is opened
right after this delete drop and before root info is updated
then the in-memory snapshot can have inconsistent results.
To prevent this retain WAL_ACT_REMOVE items till
wal_release_flushed_items

Change-Id: If7b6dce326e9fa8abc039381199728195c575acf

show more ...

a348112c25-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18325 Initialize CRITICAL_SECTION before accessing it on Windows

On windows, CRITICAL_SECTION variable should be initialized by
InitializeCriticalSection() before accessing.

MB-18325 Initialize CRITICAL_SECTION before accessing it on Windows

On windows, CRITICAL_SECTION variable should be initialized by
InitializeCriticalSection() before accessing.

Change-Id: I6c62a8f4682a42fcbdb35e1ab1cf42a1bc11d032

show more ...

27fba53d20-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18286: Support snapshot and rollback without sequence tree

Change-Id: I52e095882fb5b1741ecf4d6940fa246056890685

6ffdf64822-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

Track latencies for persisted & in-memory snapshots separately

If some application is using persisted snapshots when it should
be using in-memory snapshots instead, this change can help

Track latencies for persisted & in-memory snapshots separately

If some application is using persisted snapshots when it should
be using in-memory snapshots instead, this change can help catch
the cause of any "perf regression".

Change-Id: I0f13b77fb3fd2178fae0dc03e6c70e2b7ac07490

show more ...

34e10f8520-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-14710: MVCC WAL for Fast & Cheap Snapshots, Iterators

wal_items are shared with writer, current & all future snapshots
until wal_flush.
Mutations are inserted into KV Store's late

MB-14710: MVCC WAL for Fast & Cheap Snapshots, Iterators

wal_items are shared with writer, current & all future snapshots
until wal_flush.
Mutations are inserted into KV Store's latest mutable snapshot.
When snapshot_open is called the latest mutable snapshot of KVS
is marked as immutable and returned/cloned.
Stable iterations do merge-sort over all WAL shards but
return only the wal_item that was valid at time of snapshot.
wal_flush lazily cleans up items of closed snapshots and prevents
future snapshots from accessing items created before the flush.
Iterators simplified to be symmetric & reactive instead of proactive

Change-Id: I1a42758e28d522031f595b15ec580bc837382dd9

show more ...

315d2db418-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-18201 Incr and decr the database open counter only in fdb_open APIs

The database open counter should not be incremented in fdb_init API
because the client can explicitly invoke fdb_in

MB-18201 Incr and decr the database open counter only in fdb_open APIs

The database open counter should not be incremented in fdb_init API
because the client can explicitly invoke fdb_init before openning a
database instance. This change makes sure that the open counter should
be only updated in fdb_open APIs.

Change-Id: I399bf6908c0522ec2762e286e9407d858ded85d3

show more ...

091e379121-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Fix a compiler warning in rollback_to_wal_test()

Change-Id: Ib28d6a3d2bff151a343d5ee7aabf79f266373ef9

4ce06c3420-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

Merge remote-tracking branch 'couchbase/sherlock'

Includes the following fix:
MB-18254: WAL flush on rollback for correct snapshot open
Conflicts:
tests/functiona

Merge remote-tracking branch 'couchbase/sherlock'

Includes the following fix:
MB-18254: WAL flush on rollback for correct snapshot open
Conflicts:
tests/functional/mvcc_functional_test.cc

Change-Id: I7d00fa509d5fb4c9bb1842aad91c6a6c4927b582

show more ...


1255d75519-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18254: WAL flush on rollback for correct snapshot open

last_wal_flush_hdr_bid is not updated correctly on normal
commit of rollback. This is because last_wal_flush_hdr_bid
wrongly

MB-18254: WAL flush on rollback for correct snapshot open

last_wal_flush_hdr_bid is not updated correctly on normal
commit of rollback. This is because last_wal_flush_hdr_bid
wrongly gets updated to the last_hdr_bid even when rollback
has restored a number of items into the WAL section.

Change-Id: Ic16b0fdcc4844b0768633b3ed2d5910408d2f4bb

show more ...

2cc3d98515-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Correct the file name for iterator_uncommited_seeks() test

Change-Id: I93f74599efa4bb6b5a84cf0f74b542e3c9aeba73

f7a5b7a715-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Add a new API to return the version of a given ForestDB file

This change adds the new API fdb_get_file_version to return the string
representation of a ForestDB file version.

In

Add a new API to return the version of a given ForestDB file

This change adds the new API fdb_get_file_version to return the string
representation of a ForestDB file version.

In addition, this changes adds --header-only option to forestdb_dump tool
to print out the latest database header info including a file version.

$ ./forestdb_dump --header-only path_to_database_file

DB header info:
BID: 113 (0x71, byte offset: 462848)
DB header length: 88 bytes
DB header revision number: 5
DB file version: ForestDB v1.x format
...

Change-Id: I51e90bfb9aca3ae6c63db1292882f578f7e07cf0

show more ...

6101fcd215-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18030 Fix a bug that returning the smallest key twice

- If iterator's cursor is pointing to the smallest key which exists
in WAL only, we should handle this case separately.

MB-18030 Fix a bug that returning the smallest key twice

- If iterator's cursor is pointing to the smallest key which exists
in WAL only, we should handle this case separately.

Change-Id: I92174d6f27d6868c6394a514f481cfe867b42245

show more ...

971440aa10-Feb-2016 Tommie McAfee <tahmmee@gmail.com>

Test seeking around uncommitted data

Perform min/max/next/prev with iterator on data not yet committed

Change-Id: I42cdd83e1f0f9b410e8f8d9e524e64d2ff6fa9db

9666a83512-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Add a new API to return a string representation of ForestDB library version

A string representation of ForestDB library version is based on git-describe.

Change-Id: I834cc726d4d965b

Add a new API to return a string representation of ForestDB library version

A string representation of ForestDB library version is based on git-describe.

Change-Id: I834cc726d4d965b7ae78fa2d08e1319643d56177

show more ...

8fd3415311-Feb-2016 Tommie McAfee <tahmmee@gmail.com>

MB-18061 Don't allow closing a snapshot handle if its iterators are still active

This change makes sure that if a given snapshot handle still has active
iterators, then closing the snaps

MB-18061 Don't allow closing a snapshot handle if its iterators are still active

This change makes sure that if a given snapshot handle still has active
iterators, then closing the snapshot handle shouldn't be allowed, so that its
active iterators can be still used and properly closed later.

Change-Id: I81cef88ad2222be53081584c5d181e67a9d22dfe

show more ...

6538157f28-Dec-2015 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17004: Prune an old doc's seqnum from sequence index on WAL flush

If a key has been updated or deleted, its entry is removed from the
main index by key but the entry in the sequence t

MB-17004: Prune an old doc's seqnum from sequence index on WAL flush

If a key has been updated or deleted, its entry is removed from the
main index by key but the entry in the sequence tree/trie is not removed.
As a result, the sequence tree continues to grow indefinitely until
compaction is executed, which in turn bloats up the index node space used.
Since we do not support having duplicate values for the same key in the
same snapshot, we can prune the sequence tree to prevent file size explosion.

Change-Id: I4aa2e44261873523e41a347171fd80236d839fc0

show more ...

f1b7f68011-Feb-2016 Tommie McAfee <tahmmee@gmail.com>

removing in-repo micro-benchmark

Need to maintain benchmark tool outside of fdb reop
for the purposes of updating benchmark scenarios independently
so that we are not running differe

removing in-repo micro-benchmark

Need to maintain benchmark tool outside of fdb reop
for the purposes of updating benchmark scenarios independently
so that we are not running different versions of benchmark tool
when comparing against older versions of fdb.

maintained here:
https://github.com/couchbaselabs/fdbench-micro

Change-Id: Iaaad38118d17aa93f97d836b185690556a4dd1bf

show more ...

e767474411-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Reset allocation counter only for block reclaim condition check

- This bug makes file size grow forever without any block reclaim,
if commit size is small and the term is short.

Reset allocation counter only for block reclaim condition check

- This bug makes file size grow forever without any block reclaim,
if commit size is small and the term is short.

- Since superblock is updated for every commit, we should detach
allocation counter reset code from sb_sync_circular().

Change-Id: Id2407be4abdc0056dcd658d8879625cdbf4a047c

show more ...

35571fce08-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-17915 Update superblock for every commit

- If DB crashes without updating superblock, some commits after the last
update of superblock may be lost.

- To address this issue, u

MB-17915 Update superblock for every commit

- If DB crashes without updating superblock, some commits after the last
update of superblock may be lost.

- To address this issue, update superblock for every commit.

- We plan to address this issue in more efficient way soon.

Change-Id: I3ed8243686b22fe578ddc0e33fd215e3a5e6c266

show more ...

42b5b25004-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-17785 Keep the current tree cursor for iteration end checking

- The tree cursor for WAL entries moves before the routine that checks
if iterator reaches the end of key range.

MB-17785 Keep the current tree cursor for iteration end checking

- The tree cursor for WAL entries moves before the routine that checks
if iterator reaches the end of key range.

- If a doc right before the end of range (which implies the second
greatest or smallest doc) is deleted, then iterator incorrectly
terminates without returning the greatest/smallest doc.

Change-Id: I98402d8ea4e0ac839d16f2e368859aa2600c45f4

show more ...

c110794e04-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

Disable a warning log in checking a commit header marker

Circular block reuse can cause false alarms in reading a commit header
because a stale commit header block might have been reclai

Disable a warning log in checking a commit header marker

Circular block reuse can cause false alarms in reading a commit header
because a stale commit header block might have been reclaimed and reused for
incoming writes.

As a separate enhancement, we will implement a better approach to logging real
corruption in a commit header.

Change-Id: I91e70eea7c43887bc52aa2033743d28c46825894

show more ...

ccb6227104-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-17354 Get the max file size on-the-fly before waiting for compaction

- If test speed is too slow so that compaction is triggered before
the test routine completes the initial load, th

MB-17354 Get the max file size on-the-fly before waiting for compaction

- If test speed is too slow so that compaction is triggered before
the test routine completes the initial load, then we get the file
size of already compacted file and the test fails.

- To address this issue, we need to continuously get & update the max
file size between each fdb_set() operation.

Change-Id: I9d905accfce648873f07c482ed7c73cbc882f76e

show more ...

9483087b04-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-17355 Fix buffer overflow of stack area in rollback_forward_seqnum()

- As the test uses doc[0] ~ doc[n], we have to allocate n+1 memory
regions.

Change-Id: I272270d79ed16972b

MB-17355 Fix buffer overflow of stack area in rollback_forward_seqnum()

- As the test uses doc[0] ~ doc[n], we have to allocate n+1 memory
regions.

Change-Id: I272270d79ed16972b8213c958773834003eb673a

show more ...

0518433104-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Fix a compiler warning due to unused variables

- transaction_in_memory_snapshot_test() in mvcc_functional_test.cc
contains unused variables.

Change-Id: I0c5c678ec0b73a174921fe80

Fix a compiler warning due to unused variables

- transaction_in_memory_snapshot_test() in mvcc_functional_test.cc
contains unused variables.

Change-Id: I0c5c678ec0b73a174921fe80fa3d483e446a3bce

show more ...

8c72096204-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Fix a bug on returning remaining reusable block before compaction

- Returning logic does not work correctly so that some heap memory
regions are not freed and cause memory leak.

Fix a bug on returning remaining reusable block before compaction

- Returning logic does not work correctly so that some heap memory
regions are not freed and cause memory leak.

Change-Id: I50d42b937fad01a39f130c16ac60a0b345016f1f

show more ...

12345678910>>...53