aboutsummaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* update after review feedbackArne Juul2017-12-141-6/+3
| | | | | * combine loop exit criteria * pre-size hash table
* do de-duplication on GlobalID in dispatchArne Juul2017-12-133-9/+36
|
* Use standard locking in searchlib (pass 1).Tor Egge2017-12-121-0/+1
|
* Factor out CountDownLatch and Gate to separate header files.Tor Egge2017-12-119-9/+9
|
* Remove unused includes.Tor Egge2017-12-119-8/+2
|
* Use standard locking in searchcore (pass 2).Tor Egge2017-12-115-26/+40
|
* Use standard locking in searchcore (pass 1).Tor Egge2017-12-1111-36/+40
|
* Use standard locking in searchcore/proton/server (pass 2).Tor Egge2017-12-118-32/+42
|
* Use standard locking in searchcore/proton/server (pass 1).Tor Egge2017-12-119-27/+24
|
* Remember document type repo below SPI in proton for 10 minutes.Tor Egge2017-12-082-2/+32
|
* Merge pull request #4377 from ↵Henning Baldersheim2017-12-0724-20/+211
|\ | | | | | | | | vespa-engine/toregge/use-alternate-maxflushed-parameter-when-node-is-retired Use alternate maxflushed parameter when node is retired.
| * Use alternate maxflushed parameter (max number of flushed memory indexesTor Egge2017-12-0724-20/+211
| | | | | | | | before fusion becomes urgent) when node is retired.
* | Merge pull request #4126 from ↵Tor Brede Vekterli2017-12-072-6/+3
|\ \ | |/ |/| | | | | vespa-engine/vekterli/move-to-flex-bison-parsing-for-document-selection-rebased Use new C++ document selection parser
| * Use new C++ document selection parserTor Brede Vekterli2017-11-142-6/+3
| | | | | | | | | | | | | | | | | | | | Based on Flex/Bison and replaces old Spirit.Classic parser. New parser is pure and does not require any locking, unlike the previous implementation. This also removes parsing of the deprecated searchcolumn feature. Adds build-time dependencies on Flex and Bison.
* | Merge pull request #4357 from ↵Henning Baldersheim2017-12-051-9/+10
|\ \ | | | | | | | | | | | | vespa-engine/toregge/delay-check-of-totals-until-we-have-totals Delay check of total memory and disk gains until we have calculated the totals.
| * | Delay check of total memory and disk gains until we have calculated theTor Egge2017-12-051-9/+10
| | | | | | | | | | | | totals.
* | | Merge pull request #4355 from ↵Henning Baldersheim2017-12-054-2/+51
|\ \ \ | | | | | | | | | | | | | | | | vespa-engine/toregge/allow-more-disk-bloat-while-node-state-is-retired Allow more disk bloat while node state is retired.
| * | | Allow more disk bloat while node state is retired.Tor Egge2017-12-054-2/+51
| |/ /
* | | must not use mtf after giving it awayArne Juul2017-12-051-1/+1
| | |
* | | adjust logic after reviewArne Juul2017-12-051-8/+13
| | |
* | | fix off-by-one in coverage reportingArne Juul2017-12-051-6/+17
|/ / | | | | | | | | | | | | * using getCommittedDocIdLimit() isn't really what we want to do, it will (even in the simplest case) be 1 more than the number of indexed documents. Use getNumUsedLids(), and do an extra check to avoid division by zero.
* | Add metric for replay time of the transaction log during start-up.Geir Storli2017-12-042-1/+4
| |
* | Merge pull request #4322 from ↵Henning Baldersheim2017-12-048-46/+59
|\ \ | | | | | | | | | | | | vespa-engine/geirst/add-more-proton-resource-limit-metrics Add metrics on how far away from disk and memory resource limits prot…
| * | Add metrics on how far away from disk and memory resource limits proton is.Geir Storli2017-12-018-46/+59
| | |
* | | Split parent + container-dependency-versions from root pom.gjoranv2017-12-011-0/+1
|/ / | | | | | | | | | | - Add missing dependencies so that all provided non-yahoo jars are listed in container-dependency-versions. - Add relativePath for all child poms of parent.
* | Revert "Gjoranv/split parent2"gjoranv2017-11-301-1/+0
| |
* | Split parent + container-dependency-versions from root pom.gjoranv2017-11-301-0/+1
| | | | | | | | | | | | - Add missing dependencies so that all provided non-yahoo jars are listed in container-dependency-versions. - Add relativePath for all child poms of parent.
* | Make flushDone() wait until scheduled task has been handled.Tor Egge2017-11-293-6/+7
| | | | | | | | Simplify locking by taking strategyGuard within getSortedTargetList().
* | Adjust comments for flushDone() method (called by flush scheduler thread).Tor Egge2017-11-292-2/+2
| |
* | Add unit tests for transaction log pruning when closing flush engine andTor Egge2017-11-291-9/+55
| | | | | | | | when completing priority flush strategy.
* | Log event each time transaction log is pruned.Tor Egge2017-11-293-0/+17
| |
* | Try to prune tls when shutting down flush engine or when completingTor Egge2017-11-281-0/+3
| | | | | | | | handling of a priority flush strategy.
* | Use standard locking in FlushEngine.Tor Egge2017-11-282-38/+43
| |
* | Add bucketSpace argument to ↵Tor Egge2017-11-206-19/+21
| | | | | | | | storage::spi::PersistenceProvider::setClusterState() method.
* | Merge pull request #4113 from ↵Geir Storli2017-11-142-34/+35
|\ \ | | | | | | | | | | | | vespa-engine/toregge/use-std-mutex-in-searchcore-persistence-engine Use std::mutex in searchore PersistenceEngine.
| * | Use std::lock_guard and scopes instead of std::unique_lock.Tor Egge2017-11-141-14/+17
| | |
| * | Use std::mutex in searchore PersistenceEngine.Tor Egge2017-11-142-23/+21
| | |
* | | Merge pull request #4123 from ↵Geir Storli2017-11-148-15/+16
|\ \ \ | | | | | | | | | | | | | | | | vespa-engine/toregge/consolidate-makebucketspace-test-functions Consolidate makeBucketSpace() test functions to document module.
| * | | Consolidate makeBucketSpace() test functions to document module.Tor Egge2017-11-148-15/+16
| |/ /
* | | Merge pull request #4116 from ↵Henning Baldersheim2017-11-141-15/+6
|\ \ \ | |_|/ |/| | | | | | | | vespa-engine/balder/simplify-and-always-hold-lock-on-map Always hold the lock when accessing the handler map.
| * | Always hold the lock when accessing the handler map.Henning Baldersheim2017-11-141-15/+6
| |/ | | | | | | Simplify, and avoid naked new.
* / Use std::mutex in searchcore MatchEngine.Tor Egge2017-11-142-7/+8
|/
* Use std::lock_guard instead of std::unique_lock.Tor Egge2017-11-1310-32/+31
|
* Incorporate replay cost of a single operation in prepare restart flush strategy.Geir Storli2017-11-097-30/+72
|
* Only commit attributes once when handling a batch remove in attribute writer.Geir Storli2017-11-081-3/+41
| | | | | This avoids potential accumulation of dead memory when handling delete bucket operations while attributes are being flushed and attribute guards being held.
* Remove unused code.Geir Storli2017-11-081-6/+0
|
* Add test for batch remove on attribute writer.Geir Storli2017-11-081-6/+36
|
* Revert "Geirst/commit attributes once when handling delete bucket operations"Arnstein Ressem2017-11-072-77/+16
|
* clean up tensor engine APIHåvard Pettersen2017-11-071-7/+4
| | | | make Tensor a subclass of Value
* Only commit attributes once when handling a batch remove in attribute writer.Geir Storli2017-11-071-4/+41
| | | | | This avoids potential accumulation of dead memory when handling delete bucket operations while attributes are being flushed and attribute guards being held.