1
0
Fork 0
Commit Graph

305 Commits

Author SHA1 Message Date
Jan 6180fcfdd1 Bug fix/prevent multiple journals (#3027)
* prevent multiple journals

* fix documentation

* remove _nrDesired, as it is not used anymore
2017-08-15 23:02:08 +02:00
Frank Celler 6fe8e2de49 ensure that exit signals are still valid (#2962) 2017-08-05 17:12:33 +02:00
Jan 6b24529a75 fix small issues found by cppcheck (#2871) 2017-07-26 13:06:31 +02:00
Frank Celler a5a25754ed Feature/reduce extraction to projection (#2792)
* reduce extractions to projections

* recycle string buffers in SocketTask

* micro optimizations for mmfiles indexes

* added special lookup function for _key

* moved function into the correct file

* speed up key buffer allocations a bit

* added noexcept specifier

* correctly name variable

* explicitly move bounds

* fix and speedup from/toPersistent functions

* reuse string from ManagedDocumentResult for multiple lookups

* use move-assign

* a bit less work for single server

* speedup AQL function HASH

* single fetch optimization

* performance optimization for the case when no documents need to be returned

* make reduce-extraction-to-projection a RocksDB-only optimizer rule

* cppcheck

* try to fix compile error on MacOS

* bug fix for MacOSX

* missing namespace (in Windows compile)
2017-07-14 08:40:29 +02:00
Frank Celler 40d73d5a8b Revert "[WIP] Feature/reduce extraction to projection (#2735)"
This reverts commit 5bfcff30cd.
2017-07-12 12:52:14 +02:00
Jan 5bfcff30cd [WIP] Feature/reduce extraction to projection (#2735)
* reduce extractions to projections

* recycle string buffers in SocketTask

* micro optimizations for mmfiles indexes

* added special lookup function for _key

* moved function into the correct file

* speed up key buffer allocations a bit

* added noexcept specifier

* correctly name variable

* explicitly move bounds

* fix and speedup from/toPersistent functions

* reuse string from ManagedDocumentResult for multiple lookups

* use move-assign

* a bit less work for single server

* speedup AQL function HASH

* single fetch optimization

* performance optimization for the case when no documents need to be returned

* make reduce-extraction-to-projection a RocksDB-only optimizer rule
2017-07-12 11:22:29 +02:00
Jan ba1ef89b94 re-introduce one lock for read/write (#2686) 2017-06-28 15:11:56 +02:00
Jan a1a9334b67 unstrand SSL (#2667) 2017-06-27 00:00:34 +02:00
Frank Celler 91d08645f7 Revert/ssl fix (#2662)
* Revert "Bug fix/task locking (#2618)"

This reverts commit 0e0bf7aae3.

* Revert "fix races in SSL communication (#2591)"

This reverts commit b32db87b67.
2017-06-23 17:07:34 +02:00
Jan 0e0bf7aae3 Bug fix/task locking (#2618) 2017-06-20 16:53:20 +02:00
Jan b32db87b67 fix races in SSL communication (#2591) 2017-06-19 17:30:14 +02:00
Frank Celler 655ef21146 moved to 503 2017-06-09 22:50:52 +02:00
Kaveh Vahedipour 54e5bd92b0 fix for closed sockets while async writes were still pending 2017-06-09 21:13:23 +02:00
Frank Celler cf89496faa shutdown will close sockets immediately (#2540) 2017-06-09 16:08:54 +02:00
jsteemann b334a6132c Merge branch 'devel' of https://github.com/arangodb/arangodb into devel 2017-06-09 14:51:20 +02:00
jsteemann 8add368a78 merged shutdown and close 2017-06-09 14:51:11 +02:00
Kaveh Vahedipour 6348945711 revert ssl handshake timeout 2017-06-09 14:04:59 +02:00
jsteemann 21ac01fba6 remove unused functions 2017-06-09 12:29:14 +02:00
Kaveh Vahedipour 3f0b4ad4ee investigating ssl handshake bug 2017-06-09 11:58:45 +02:00
jsteemann 9c7df7e5b9 make arangod abort again when address is already in use 2017-05-17 11:40:40 +02:00
jsteemann cbba71bb00 change feature order around 2017-05-10 14:29:20 +02:00
jsteemann 217d41f6f5 fix shutdown races 2017-05-09 10:24:40 +02:00
jsteemann 05e0c08824 there still seems to be a race between
- WorkMonitorThread asynchronously deleting handlers, and handlers accessing the io service
- the io service being deleted by the scheduler
2017-05-09 02:03:59 +02:00
jsteemann fc0170f2a8 re-enable shared_from_this 2017-05-08 22:06:03 +02:00
jsteemann 129abe71b5 mmh. did the previous change break the cluster? 2017-05-08 18:39:47 +02:00
jsteemann 788fc07427 fix race 2017-05-08 17:57:42 +02:00
jsteemann 2cffe15f0d fix wrong assertion 2017-05-08 11:23:49 +02:00
jsteemann ac7e0a225f fixed issues found by scan tool 2017-05-05 13:51:23 +02:00
jsteemann edeeef8b56 re-enabled signal handlers 2017-05-05 13:09:59 +02:00
jsteemann d47a506bcc slightly improve error messages 2017-05-04 23:17:26 +02:00
jsteemann 0d964bc704 try to work around races 2017-05-04 13:10:53 +02:00
jsteemann b9e87d2619 fixed a shutdown race 2017-05-03 15:04:31 +02:00
jsteemann 1d22f7bb61 yield on shutdown 2017-04-28 10:06:27 +02:00
jsteemann aa521d5412 better error messages 2017-04-27 14:47:19 +02:00
jsteemann 9161d48788 changed option description 2017-04-26 13:37:40 +02:00
Simon Grätzer a8285641fb Merge branch 'devel' of https://github.com/arangodb/arangodb into devel 2017-04-26 11:30:44 +02:00
Simon Grätzer 6472c4928d Various warnings 2017-04-26 11:26:08 +02:00
Frank Celler ab7bcfe224 --server.maximal-queue-size is now an absolute maximum 2017-04-26 11:20:56 +02:00
Simon Grätzer d228a94462 Merge branch 'devel' of https://github.com/arangodb/arangodb into devel
# Conflicts:
#	arangod/Replication/InitialSyncer.cpp
2017-04-26 11:06:09 +02:00
jsteemann 09e3a19a5b try to fix shutdown issue 2017-04-25 14:41:04 +02:00
Frank Celler 45690bbbdd fixed queue size 2017-04-24 18:47:44 +02:00
Frank Celler 1a4675dfbf added queue size to statistics 2017-04-24 18:47:44 +02:00
Frank Celler d783d4ecae added queue time and request tracing with timings 2017-04-24 18:47:44 +02:00
jsteemann fcc61115d8 try to fix feature startup order (Scheduler vs. CacheManager vs. DatabaseFeature) 2017-04-24 13:17:21 +02:00
jsteemann c5854d050b fix shutdown issue, modernize thread creation a bit 2017-04-19 16:57:53 +02:00
jsteemann e5ba517ecf fixed shutdown issue 2017-04-05 10:43:05 +02:00
jsteemann d37ef3b2a7 don't make Scheduler depend on Database again. needs to be fixed properly 2017-04-05 08:35:44 +02:00
jsteemann ad287a406c fix a shutdown error 2017-04-04 17:23:23 +02:00
Simon Grätzer 664f45d029 Fixed counter 2017-04-03 15:16:14 +02:00
jsteemann 058e6002d3 try to fix shutdown races 2017-03-22 08:31:49 +01:00
Frank Celler e997b791a8 fixed lost HTTP requests 2017-03-21 22:51:49 +01:00
Jan Christoph Uhde f21f88c0d4 Merge branch 'devel' of https://github.com/arangodb/arangodb into devel
* 'devel' of https://github.com/arangodb/arangodb:
  Pregel: fixed catch test
  unify case in error messages
  move collection keys into engine
2017-03-21 15:28:31 +01:00
Jan Christoph Uhde cb3dd9683f fix logtopic 2017-03-21 15:27:52 +01:00
jsteemann d43a4073ad unify case in error messages 2017-03-21 14:24:47 +01:00
Jan Christoph Uhde 5a60a52db9 Merge branch 'devel' of https://github.com/arangodb/arangodb into devel
* 'devel' of https://github.com/arangodb/arangodb:
  move replication-dump into engine
  try to fix shutdown races
  changelog
  ui - fixed some event issues
  moved export into engine
  move attributes into engine
2017-03-21 14:10:43 +01:00
Jan Christoph Uhde 1c9ad06aa6 Avoid name resolution if given host is a valid address 2017-03-21 14:07:01 +01:00
jsteemann 93423ba273 try to fix shutdown races 2017-03-21 12:59:46 +01:00
jsteemann 9349005191 wait longer before scheduler shutdown 2017-03-20 11:05:24 +01:00
Frank Celler 31de362d84 switch to uint64 for hidden options --server.minimal-threads and --server.maximal-threads 2017-03-17 12:34:47 +01:00
Max Neunhoeffer 428b6aa67f Port thread fixes from 3.1 to devel. 2017-03-16 13:53:40 +01:00
jsteemann 4727fd0f4b remove unused file 2017-03-13 17:32:13 +01:00
jsteemann 695d215b0c ensure that JobGuards are started and ended in the same thread 2017-02-27 22:41:58 +01:00
jsteemann 4005a15697 added some assertions 2017-02-27 17:23:41 +01:00
Frank Celler 4a3fdf6351 raised default hard limit on threads for very small to 64 2017-02-23 09:08:39 +01:00
Frank Celler a38f19a84c added locking to avoid races within boost asio 2017-02-22 14:51:48 +01:00
Frank Celler b2b9f377c1 safe guard 2017-02-22 09:02:49 +01:00
jsteemann af17235242 fix too early closing of sockets 2017-02-16 15:42:13 +01:00
jsteemann d024a6d00a remove logging for non-topics 2017-02-10 09:32:50 +01:00
Frank Celler 0da95481ea added locks for keep-alive 2017-02-08 09:41:18 +01:00
Frank Celler b3b02b10d3 fixed nested JobGuards 2017-02-07 14:35:24 +01:00
Frank Celler 00720cdd29 ignore the result of processAll, try to read more async 2017-02-07 13:33:03 +01:00
Frank Celler e3fe51faa8 better threads working handling 2017-02-06 16:32:55 +01:00
jsteemann ab11410911 Merge branch 'devel' of https://github.com/arangodb/arangodb into engine-api 2017-02-06 13:05:11 +01:00
jsteemann 945edacf3f move connection info 2017-02-03 17:43:57 +01:00
jsteemann 952b9c00c2 Merge branch 'devel' of https://github.com/arangodb/arangodb into engine-api 2017-02-03 10:43:56 +01:00
Frank Celler 7e2e8bfee0 reset counter as well 2017-02-02 16:00:03 +01:00
jsteemann 072553d3f3 don't throw in dtor 2017-02-02 10:21:15 +01:00
jsteemann fa917937c4 do not use namespaces in header files 2017-02-01 13:41:31 +01:00
Max Neunhoeffer c4d5d982c7 Fix bugs of threads not announcing that they are busy. 2017-01-31 14:07:54 +01:00
Dan Larkin 5b736dc319 Changed index-filling to dispatch batch insertion tasks to scheduler, obsoleted index-threads option. 2017-01-30 13:05:10 -05:00
jsteemann 43fb70a693 fix double-free issue 2017-01-30 15:32:06 +01:00
Frank Celler e0c00c6de6 improved statistics 2017-01-26 23:43:27 +01:00
jsteemann 34a110a08b mini fixes 2017-01-26 10:49:42 +01:00
Frank Celler fa114470f5 fixes for statistics 2017-01-25 22:31:00 +01:00
Max Neunhoeffer b83df2fbcd Typo in comment. 2017-01-25 15:38:43 +01:00
jsteemann 41e3268ccd clean up job queues on shutdown 2017-01-25 10:26:30 +01:00
Max Neunhoeffer c1428e81cd Try to fix a shutdown bug in Scheduler. 2017-01-24 16:47:23 +01:00
Jan Christoph Uhde d92101d74c clean up code 2017-01-12 10:28:48 +01:00
Jan Christoph Uhde 4858469bfa try to fix a bug in server that surfaced on heavy load 2017-01-11 15:27:53 +01:00
Jan Christoph Uhde 62eda019c2 Revert "switch immediately to async writes when socket would block during write"
This reverts commit 96067a9315.

The change resulted in a stackoverflow instead of a "hanging" connection
we might need to use something like TCP Keep Alive to handle
disconnected clients. The bug we are currently working on might not be
fixed by TCP Keep Alive as we are in a heavy load situation. Still the
socket should not block during write. It could do so if the client does
not receive the data form the network and the server side networkbuffer
runs full at some point. There might be other situations we are not
aware of.
2017-01-11 10:45:18 +01:00
Jan Christoph Uhde 96067a9315 switch immediately to async writes when socket would block during write 2017-01-10 13:47:31 +01:00
jsteemann a92b49d46e fix races in VST response handling 2017-01-09 17:06:33 +01:00
Max Neunhoeffer 1a30813919 Fix compilation. 2016-12-22 16:00:14 +01:00
Max Neunhoeffer 65c662b3f7 Enforce maximal number of threads. 2016-12-22 15:53:54 +01:00
Max Neunhoeffer e08d8cc1f8 Add a debug message about detected processors. 2016-12-20 14:46:43 +01:00
jsteemann 4f10a6fb9c experimental micro optimizations 2016-12-16 15:52:16 +01:00
jsteemann 3e6064f772 preliminary fix for issue #2218 2016-12-15 15:45:33 +01:00
jsteemann 9b42e9cee6 some improvements for memory management
added more memory diagnostics for memory usage:
- collection.figures() now returns a "readCache" attribute which contains the
  number of and memory usage of entries in the document revisions cache for the
  collection, and a "revisions" attribute with the number and memory usage
  of entries in the storage engine's revision lookup table for the collection
- the default value for --database.revision-cache-target-size was changed from
  75% of RAM to 30% of RAM
2016-12-15 15:30:28 +01:00
jsteemann b96897e296 fixed some issues detected by coverity build tool 2016-12-12 15:41:54 +01:00
Frank Celler 93eb443366 removed left-over merge conflict 2016-12-05 16:36:27 +01:00