]> git-server-git.apps.pok.os.sepia.ceph.com Git - rocksdb.git/log
rocksdb.git
10 years agoChange RocksDB version to 4.1 rocksdb-4.1
Yueh-Hsuan Chiang [Thu, 8 Oct 2015 18:15:18 +0000 (11:15 -0700)]
Change RocksDB version to 4.1

Summary: Change RocksDB version to 4.1

Test Plan: no code change.

Reviewers: sdong, anthony, IslamAbdelRahman, igor

Reviewed By: igor

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D48387

10 years agoEnable crash CI jobs
krad [Thu, 8 Oct 2015 17:13:01 +0000 (10:13 -0700)]
Enable crash CI jobs

Summary: Enabling CI crash test jobs

Test Plan: Manual testing

Reviewers:

CC: leveldb@
Task ID: #8635140

Blame Rev:

10 years agocompaction_filter.h cleanup
Igor Canadi [Thu, 8 Oct 2015 16:32:50 +0000 (09:32 -0700)]
compaction_filter.h cleanup

Summary:
Two changes:
1. remove *V2 filter stuff. we deprecated that a while ago
2. clarify what happens when user sets max_subcompactions to bigger than 1

Test Plan: none

Reviewers: yhchiang, sdong

Reviewed By: sdong

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D47871

10 years agoBytes read/written from cache statistics
Islam AbdelRahman [Wed, 7 Oct 2015 22:17:20 +0000 (15:17 -0700)]
Bytes read/written from cache statistics

Summary: Add 2 new counters BLOCK_CACHE_BYTES_WRITE, BLOCK_CACHE_BYTES_READ to keep track of how many bytes were written to the cache and how many bytes that we read from cache

Test Plan: make check

Reviewers: sdong, yhchiang, igor

Reviewed By: igor

Subscribers: dhruba

Differential Revision: https://reviews.facebook.net/D48195

10 years agoCreate Makefile target unity_test
Igor Canadi [Wed, 7 Oct 2015 21:46:18 +0000 (14:46 -0700)]
Create Makefile target unity_test

Summary: unity_test will compile db_test with unity.a instead of librocksdb.a. This will test both the compilation and some small amount of runtime.

Test Plan: This is a test :)

Reviewers: kradhakrishnan

Reviewed By: kradhakrishnan

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D48297

10 years agobloom hit/miss stats for SST and memtable
dyniusz [Wed, 7 Oct 2015 18:23:20 +0000 (11:23 -0700)]
bloom hit/miss stats for SST and memtable

Summary:
hit and miss bloom filter stats for memtable and SST
stats added to perf_context struct
key matches and prefix matches combined into one stat

Test Plan: unit test veryfing the functionality added, see BloomStatsTest in db_test.cc for details

Reviewers: yhchiang, igor, sdong

Reviewed By: sdong

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D47859

10 years agoFix compile error on platforms without fallocate()
Igor Canadi [Wed, 7 Oct 2015 18:02:23 +0000 (11:02 -0700)]
Fix compile error on platforms without fallocate()

Summary:
If a platform doesn't have ROCKSDB_FALLOCATE_PRESENT, then compiler complains:

util/env_posix.cc:354:8: error: private field 'allow_fallocate_' is not used [-Werror,-Wunused-private-field]

This was caught by travis.

Test Plan: compiles with ROCKSDB_FALLOCATE_PRESENT.

Reviewers: sdong

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D48327

10 years agoFix compile failure on Travis
Igor Canadi [Wed, 7 Oct 2015 17:17:47 +0000 (10:17 -0700)]
Fix compile failure on Travis

Summary:
Travis is complaining against using {} to initialize KVMap: https://travis-ci.org/facebook/rocksdb/jobs/84132600

      db/compaction_job_test.cc:526:26: error: chosen constructor is explicit in copy-initialization
        RunCompaction({files}, {});

This diff should fix it

Test Plan: travis

Reviewers: sdong

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D48309

10 years agoAdded boolean variable to guard fallocate() calls
Lakshmi Narayanan [Wed, 7 Oct 2015 17:04:05 +0000 (10:04 -0700)]
Added boolean variable to guard fallocate() calls

Summary:
Added boolean variable to guard fallocate() calls.
Set to false to prevent space leaks when tests fail.

Test Plan:
Compliles
Set to false and ran log device tests

Reviewers: sdong, lovro, igor

Reviewed By: igor

Subscribers: dhruba

Differential Revision: https://reviews.facebook.net/D48027

10 years agoTravis shouldn't fail when clang-format suggests improvements
Igor Canadi [Wed, 7 Oct 2015 16:41:21 +0000 (09:41 -0700)]
Travis shouldn't fail when clang-format suggests improvements

Summary: This diff avoids failing travis test when clang-format fails. We should keep the test green even if clang-format fails. It would be good to look at travis output for pull requests to make sure they're following the good code style. However, we broke travis test 3 times today because of formatting issues. We should keep travis test green if the only thing wrong is the formatting.

Test Plan: none

Reviewers: sdong, yhchiang, anthony

Reviewed By: anthony

Subscribers: dhruba, leveldb

Differential Revision: https://reviews.facebook.net/D48291