Commit graph

5749 commits

Author SHA1 Message Date
Rich Prohaska
5dc1ef2f4f #56 simply launch of lock escalator and write a slightly better test 2013-10-08 14:52:34 -04:00
Zardosht Kasheff
27f5720d4a refs #82, fix CMakeLists.txt 2013-10-07 10:40:40 -04:00
Zardosht Kasheff
4f74e29926 refs #82, remove checkpoint_1.cc for TokuDB 7.1.0, may bring it back fixed later. That is still open 2013-10-07 10:35:05 -04:00
Zardosht Kasheff
548bc646b2 refs #78 for minicrons with a period of longer than one second (checkpoints),
change minicron to count the period from the beginning of
the callback's execution and not the end. For checkpoints, this makes them
reliably start every 60 seconds.
2013-10-07 10:17:03 -04:00
Zardosht Kasheff
6e60da0886 refs #84, remove some shared variables from logger 2013-10-07 10:16:03 -04:00
rich prohaska
d067419ae9 tokutek/ft-index#80 tokutek/ft-engine#94 impose an upper bound on loader memory reservations 2013-10-07 08:35:52 -04:00
Rich Prohaska
2796c2d35c tokutek/ft-index#76 add US Patent 8,489,638 2013-10-04 16:49:53 -04:00
Zardosht Kasheff
77903f0860 refs #61, fix locking bug, add write list lock before reintegrating PAIRs into the cachetable on a open 2013-10-04 11:34:49 -04:00
Zardosht Kasheff
4cba362f5c refs #61, fix some tests, remove checkpoint_callback.cc, as it is not an interesting test, still need t ofix checkpoint_1.cc 2013-10-03 16:35:03 -04:00
Zardosht Kasheff
4d234861e1 closes #77, remove tabs from memory.cc 2013-10-02 12:01:14 -04:00
Rich Prohaska
bbee1fec1f #56 run lock escalation on a background thread 2013-10-02 10:45:25 -04:00
Zardosht Kasheff
64cdcbc261 refs #61,
- have closed cachefiles not immedietely free pairs, but set them to the side
 - leave freeing of pairs to the evictor and/or shutdown
 - should a cachefile be reopened before all pairs are freed,
   the pairs belonging to that cachefile are reintegrated into the cachetable
2013-10-02 09:53:00 -04:00
Rich Prohaska
013bea408f #59 get test_lock_timeout_callback to work with valgrind. change the type of a sync_fetch_and_add from bool to int 2013-10-01 10:22:03 -04:00
Zardosht Kasheff
bdd2260da3 This checkin was meant for a branch. Undoing
Revert "stuff"

This reverts commit 2423c9d0ef.
2013-09-26 11:33:47 -04:00
Zardosht Kasheff
51495a5493 stuff 2013-09-26 11:27:25 -04:00
Zardosht Kasheff
fbadc9f864 refs #61, remove test helper function toku_cachefile_flush,
fix tests that use it to not need it
2013-09-26 11:10:12 -04:00
Zardosht Kasheff
b9b60dd986 refs #61, fix optimized compilation that has TOKU_DEBUG_PARANOID off 2013-09-25 18:10:57 -04:00
Zardosht Kasheff
4e69563663 refs #61, several code simplifications:
- break up cachetable_flush_cachefile into more digestable functions,
 - decouple hash_id from filenum
 - break up close_userdata into close_userdata and free_userdata
2013-09-25 17:49:05 -04:00
Zardosht Kasheff
e68aba3046 refs #61, clean up code relating to cachefiles, expand the
cachefiles_list class and move some functionality in there.
2013-09-25 15:23:03 -04:00
Yoni Fogel
bf4c67f20d refs #46, LOTS of refactoring done.
Isolate mempool and OMT into a new class, bndata.
Remove key from the leafentry.
2013-09-25 11:07:30 -04:00
Rich Prohaska
1a75e45c41 #56 benchmark that demos lock escalation stalls 2013-09-19 17:21:29 -04:00
John Esmet
17267901ef Merge branch 'master' of github.com:Tokutek/ft-index 2013-09-19 09:20:43 -04:00
Rich Prohaska
f9541b70af #69 add long tail counts to global status 2013-09-18 14:52:43 -04:00
John Esmet
c3089da11b fixes #71 fix a test bug by caching the txnid instead of trying to use a
txn object after it commits or aborts
2013-09-18 13:41:00 -04:00
Rich Prohaska
b846739abf #69 measure long tail lock tree and cache table stalls 2013-09-18 13:29:05 -04:00
Rich Prohaska
0e689c58f9 Merge branch 'master' of github.com:Tokutek/ft-index 2013-09-18 09:51:10 -04:00
Rich Prohaska
018705472a #50 add long tail counts to global status 2013-09-18 09:51:02 -04:00
John Esmet
8a9faf4d60 Merge branch 'master' of github.com:Tokutek/ft-index 2013-09-18 09:49:43 -04:00
John Esmet
957bd4e939 Merge branch 'bugs/70' 2013-09-18 09:48:03 -04:00
John Esmet
e9be482ebf Fix this test - did not get built or tested properly on my end due to
BUILD_TESTING=Off in the cmake config.
2013-09-18 09:47:18 -04:00
Rich Prohaska
81e63d1dfd #50 add long tail counts to global status 2013-09-18 08:48:39 -04:00
John Esmet
e0e1cc9984 Merge branch 'bugs/70' 2013-09-18 00:19:25 -04:00
John Esmet
387523dc63 refs #70 Add a test to the range buffer that verifies the buf grows
properly after a small append into a large append
2013-09-18 00:18:57 -04:00
John Esmet
4fd61b9d8e Merge branch 'bugs/70' 2013-09-17 23:55:54 -04:00
John Esmet
ae87b251c8 fixed memory allocation error in range_buffer #70
fixes #70
2013-09-17 23:55:37 -04:00
Rich Prohaska
4cab5fd9c1 refs #59 turn off bdb builds for tokudb lock tree tests 2013-09-17 11:40:17 -04:00
Rich Prohaska
86e815082c refs #64 fix cpp guard 2013-09-17 11:39:26 -04:00
Rich Prohaska
4e3aa725f1 refs #64 fix clang on linux problems 2013-09-17 11:30:40 -04:00
John Esmet
e36ddbd5e9 Increase the lock timeout and sleep time to make this test less
timing-dependent (though it still is)
2013-09-14 10:48:35 -04:00
John Esmet
3e85b387f4 refs #59 Add the locktree visualization APIs, new accessors in the DB and DB_TXN,
and a new operation in test_stress0 for stress testing coverage
2013-09-13 11:10:49 -04:00
Leif Walsh
f8ae1ecfe8 remove -u option from cp, fixes #66 2013-09-12 14:41:44 -04:00
Leif Walsh
25dfdec3f9 fixed required cmake version #65
fixes #65
2013-09-12 09:00:47 -04:00
Rich Prohaska
7cc6daa642 refs #62 fix how tdb_logprint formats XID pairs 2013-09-08 11:32:48 -04:00
Rich Prohaska
50aef9b7d9 #50 count lock tree timeouts 2013-08-21 16:10:43 -04:00
zkasheff
f40b466bbe refs #36, have the FIFO realloc its buffer on resize, as opposed to malloc and memcpy 2013-08-21 15:29:30 -04:00
Zardosht Kasheff
a34734ccbc refs #54, improve the performance of hot indexing. This change does two things:
- gets indexer to run in reverse, that is, start at the end and run to beginning
 - refines locking a bit. An estimate of the position of the hot indexer is stored,
   that is cheap to look at. Threads that use this estimate with a mutex either do
   only a quick comparison or set it to a new value. Threads doing writes (with XXX_multiple calls)
   will check their position with respect to the estimate, and if they see the hot indexer
   is already past where they will modify, they don't grab the more expensive indexer
   lock. For insertion workloads that go to the end of the main dictionary of a table/collection,
   this check should practically always pass.
2013-08-21 15:10:05 -04:00
Rich Prohaska
b23b43a0d9 #50 count long tail events like long fsyncs, checkpoints, and lock tree waits 2013-08-20 13:36:45 -04:00
zkasheff
146344e9bb fixes #55, fix test 2013-08-20 11:31:13 -04:00
Rich Prohaska
a189ec77a1 Merge branch 'bdb-compile-fix' of github.com:Tokutek/ft-index into bdb-compile-fix 2013-08-19 14:56:51 -04:00
Rich Prohaska
4aa8376d08 #31 remove more tokudb only tests 2013-08-19 14:56:27 -04:00