History log of /6.0.3/forestdb/ (Results 176 - 200 of 1302)
Revision (<<< Hide revision tags) (Show revision tags >>>)Date Author Comments
(<<< Hide modified files)
(Show modified files >>>)
ca98b30114-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18674 Tolerate false alarms for ForestDB v1.x files

- If a document length structure is written over two blocks, we may
get a false alarm if the rest of the first block was simply ski

MB-18674 Tolerate false alarms for ForestDB v1.x files

- If a document length structure is written over two blocks, we may
get a false alarm if the rest of the first block was simply skipped
and the next document began from the beginning of the second block.

- There can be also garbage (allocated but not written) blocks in old
files. We should tolerate them instead of compaction failure.

Change-Id: I5157bde50b45a802d81266acc99b10ed6d640870

show more ...

a308683411-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18158 Use old file's version info during the migration of KVS header info

- When we read KVS header info from the old file, we should use the old
file's version info, not the latest m

MB-18158 Use old file's version info during the migration of KVS header info

- When we read KVS header info from the old file, we should use the old
file's version info, not the latest magic.

- As Sherlock branch uses the oldest magic number (MAGIC_000), we get
rid of the code block that denies opening old files.

Change-Id: I322309c0993e2d1fada9fdfb81b79be5a9c52630

show more ...

9d0f23a210-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18267 Take another superblock if DB header is corrupted

- We should take another (second latest) superblock when the DB
header pointed to by the latest superblock is corrupted althoug

MB-18267 Take another superblock if DB header is corrupted

- We should take another (second latest) superblock when the DB
header pointed to by the latest superblock is corrupted although
the superblock itself is not broken.

Change-Id: I0f15a07d0f2a69984773b809e2470efceeba0b13

show more ...

912efc0909-Mar-2016 Will Gardner <will.gardner@couchbase.com>

Add support for non-mobile ARM linux platforms

Spinlocks should be initialised on ARM. This changes adds the
necessary configuration to allow this. It also includes some of
the confi

Add support for non-mobile ARM linux platforms

Spinlocks should be initialised on ARM. This changes adds the
necessary configuration to allow this. It also includes some of
the configuration for e.g. data sizes from the Android configuration
which reduces some of the warnings

Change-Id: Ic1948ac6d0fabe5d2025c9681558ea60abf2f0fd

show more ...

f3ba0e3c09-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18539 Do not allow get_byoffset() for transaction marker

- A transaction marker contains an empty key field so that it may
causes crash if get_byoffset() is issued on the offset of th

MB-18539 Do not allow get_byoffset() for transaction marker

- A transaction marker contains an empty key field so that it may
causes crash if get_byoffset() is issued on the offset of the marker.

Change-Id: Iac9dbd541fb0897afd82d24a9657cd87341da1c0

show more ...

e97ca92809-Mar-2016 Chiyoung Seo <chiyoung@couchbase.com>

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


48e91c2009-Mar-2016 Chiyoung Seo <chiyoung@couchbase.com>

Merge "build anomaly test with objects for faster build time"


cc687f0c09-Mar-2016 Chiyoung Seo <chiyoung@couchbase.com>

Merge "Removing unused parameters in src/wal.cc"


c29d383c09-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

Do not cache DB header in blockcache for performance

DB Header is read infrequently for durable snapshots.
For these it is ok to read the header from disk so
more doc & index blocks

Do not cache DB header in blockcache for performance

DB Header is read infrequently for durable snapshots.
For these it is ok to read the header from disk so
more doc & index blocks are cached for readers.

Change-Id: I4011fe322d500dbbec928427f050a9fd36806ddb

show more ...

42d7c11509-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

build anomaly test with objects for faster build time

Also fix mvcc_tests to avoid block re-use to test blockcache behavior

Change-Id: Idd3e086a06b042cbbe872092a931b1345bb57951

c7f1c12109-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

Do not fetch the latest header BID on persistent snapshot creation

- On snapshot creation, we don't need to scan until we meet the latest
header. Just stop at the header corresponding to

Do not fetch the latest header BID on persistent snapshot creation

- On snapshot creation, we don't need to scan until we meet the latest
header. Just stop at the header corresponding to the snapshot.

- This change does not affect any DB semantics; just removing an
inefficiency.

Change-Id: I02cc4af2ad986df2a483876c561268bd994ff12a

show more ...

baeceffb08-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

Removing unused parameters in src/wal.cc

This will address warnings during compilation.

Change-Id: Ibb4b9b0d6b3c27ddf06fa9cf0fe0a6d509d437e0

5787cd6908-Mar-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-14710: Enable stats for in-memory snapshots

Change-Id: I6ba419a171443db55189340ad60d2c6d08902619

fe28a7d107-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

MB-18539: Add unittest to get doc using invalid offset

Covers 4 scenarios:
- get by offset over empty file
- get by incorrect offset that points to some data block
- get by old o

MB-18539: Add unittest to get doc using invalid offset

Covers 4 scenarios:
- get by offset over empty file
- get by incorrect offset that points to some data block
- get by old offset over a compacted file
- get by incorrect offset that points to an index block

Change-Id: I416839193bf717f493945f38c47841a1a710ef8e

show more ...

4e79303308-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18567 Insert a dummy block into buffer cache before use it

- When a block is allocated but not written back into the DB file
(this can happen due to various reasons), both block cache

MB-18567 Insert a dummy block into buffer cache before use it

- When a block is allocated but not written back into the DB file
(this can happen due to various reasons), both block cache and DB
file do not contain the block so that garbage data is read if we
call filemgr_read().

- If the block marker of the garbage block on the file side is
accidentally set to document block marker, various kinds of
unexpected behavior can happen.

- To avoid this issue, we insert a dummy block including a specific
block marker into buffer cache once it is allocated.

- It happens on the B+tree index blocks only so that we don't do the
same work for document blocks due to performance issue.

Change-Id: Id6a97892491ebae81d22b37d114335d04074024b

show more ...

a44835e508-Mar-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-18571 Finalize and flush dirty updates before creating a new KV instance

- Since _fdb_kvs_create() calls filemgr_commit() and appends a new DB
header, all previous dirty updates shoul

MB-18571 Finalize and flush dirty updates before creating a new KV instance

- Since _fdb_kvs_create() calls filemgr_commit() and appends a new DB
header, all previous dirty updates should be flushed before the
commit.

Change-Id: Iea79e3e25ff6e16cb8f82541608a0be2b04307de

show more ...

Revision tags: v1.0
cce989f824-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-18286: disable seqtree by default

Sequence tree are only required when iterating
by sequence number or fetching by sequence number
It incurs a non-trivial overhead. Disabling it

MB-18286: disable seqtree by default

Sequence tree are only required when iterating
by sequence number or fetching by sequence number
It incurs a non-trivial overhead. Disabling it
for users who do not use above 2 features boosts
space and time performance of forestdb sometimes
by over 15%. If an app requires sequence iteration
or fetch, it can be enabled via config.seqtree_opt

Change-Id: I1dade2ef37ec11cec39f25a1736404e72404a505

show more ...

c1ea545123-Feb-2016 Sundar Sridharan <sundar.sridharan@gmail.com>

MB-17239: Track Latency stats for iterator apis

Change-Id: I49de5a51dde5b64f71e7aba79018c7f7ed5e9e90

ac2374b418-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-14710 Make in-memory copy of dirty index blocks during WAL flushing

- Maintain snapshots for dirty index blocks for each dirty WAL flushing.

- As index blocks in DB file are not

MB-14710 Make in-memory copy of dirty index blocks during WAL flushing

- Maintain snapshots for dirty index blocks for each dirty WAL flushing.

- As index blocks in DB file are not overwritten by on-going WAL
flushing, we don't need to grab filemgr_mutex anymore for snapshot
or iterator initizliation.

Change-Id: I35d2183d426c088cd0fbc9fd7e0d5ebc4c8a03eb

show more ...

671d9fba19-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17757 Return file system's error code to the client upon operation failure

A file system's error can happen while performing various file operations and
is captured in errno in linux

MB-17757 Return file system's error code to the client upon operation failure

A file system's error can happen while performing various file operations and
is captured in errno in linux and _get_errno() or GetLastError on Windows.
These errors are logged through the log callback from the client, but not
returned to the client as part of a response.

To address the above issue, this commit adds a new set of error codes that
correspond to potential errno codes that can happen while performing file
operations.

Change-Id: I84fe2edd281689570798e30cee5afe054a972a90

show more ...

ce30d60a17-Feb-2016 Chiyoung Seo <chiyoung.seo@gmail.com>

MB-17670 Improve error handling in doc I/O and its callers

This change adapts the return type of doc_io_read APIs to return appropriate
error codes to the callers, so that the callers ca

MB-17670 Improve error handling in doc I/O and its callers

This change adapts the return type of doc_io_read APIs to return appropriate
error codes to the callers, so that the callers can handle those errors
properly and return them to the clients.

Change-Id: I0665fe47150e0d690db7ffc38a202c70a422cd07

show more ...

e4935a2027-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-16594 Deny reading a very old DB file

- Since DB files that are based on the initial magic number (renamed
to FILEMGR_MAGIC_000) don't contain any version info for identifying
var

MB-16594 Deny reading a very old DB file

- Since DB files that are based on the initial magic number (renamed
to FILEMGR_MAGIC_000) don't contain any version info for identifying
various DB formats, we cannot verify if its DB header is correct or
not.

Change-Id: I26f83b9ab172a11c15be845f325d53961f1d066d

show more ...

67111c2926-Feb-2016 Jung-Sang Ahn <jungsang.ahn@gmail.com>

MB-16626 Return failure instead of crashing in HB+trie

- The meta structure of B+tree node in an old DB file (before commit
'Fix potential problem caused by chunk size', committed on Jan

MB-16626 Return failure instead of crashing in HB+trie

- The meta structure of B+tree node in an old DB file (before commit
'Fix potential problem caused by chunk size', committed on Jan 17, 2015,
https://github.com/couchbase/forestdb/commit/47025c39115e69145c8423b1840bf78c9eb2898f)
is differ to the current structure, so that it causes crash during HB+trie
traversal.

- Instead of crash, we add a new error code for HB+trie and return it
without terminating the process.

Change-Id: Ic234d22a1cb70262e642e0257c508a30c660c44d

show more ...

f5882dc903-Mar-2016 Chiyoung Seo <chiyoung@couchbase.com>

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


7decd25a02-Mar-2016 abhinavdangeti <abhinav@couchbase.com>

[1] Adding new module testsuite for stale block reuse

This testsuite will contain functional tests that are
to cover stale block reuse.

Change-Id: I80f7adcf0b775b5db32debccb5953

[1] Adding new module testsuite for stale block reuse

This testsuite will contain functional tests that are
to cover stale block reuse.

Change-Id: I80f7adcf0b775b5db32debccb595378b4630dfc6

show more ...

12345678910>>...53