aboutsummaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
...
| * implement createFilterSearch in whitelist blueprintArne Juul2020-05-281-2/+5
| |
* | Use make_unique and make_shared instead of new.Geir Storli2020-05-2919-35/+33
|/
* Propagate provider tombstone metadata to internal SPI GetResultTor Brede Vekterli2020-05-262-1/+2
|
* Set default remove block rate for lid space compaction to 100 removes / sec.Geir Storli2020-05-252-2/+2
|
* Merge pull request #13331 from ↵Tor Egge2020-05-207-37/+37
|\ | | | | | | | | vespa-engine/geirst/prepare-for-matched-elements-only-on-primitive-types Prepare to support matched-elements-only for arrays and weighted sets…
| * Prepare to support matched-elements-only for arrays and weighted sets of ↵Geir Storli2020-05-207-37/+37
| | | | | | | | | | | | | | primitive types. This renames StructFieldMapper to MatchingElementsFields with added support for basic fields. Also some terminology is aligned.
* | added commentHåvard Pettersen2020-05-201-0/+2
|/
* Merge pull request #13307 from ↵Henning Baldersheim2020-05-193-0/+7
|\ | | | | | | | | vespa-engine/havardpe/let-compile-cache-use-shared-executor let compile cache use shared proton executor
| * let compile cache use shared proton executorHåvard Pettersen2020-05-193-0/+7
| |
* | - Handle more than 64k hits in the element vector.Henning Baldersheim2020-05-181-6/+6
|/ | | | | - Avoid computing all vectors in full separately and instead do an incremental inline merge with. - Also avoid requiring the searchiterator aspect on the wrappers.
* Run initializer task with highest transient memory usage first.Tor Egge2020-05-185-2/+39
|
* Merge pull request #13255 from ↵Arne H Juul2020-05-186-4/+40
|\ | | | | | | | | vespa-engine/arnej/use-white-list-bitvector-as-global-filter add a special API to get a whitelist bitvector
| * minor fixes after reviewArne Juul2020-05-143-9/+14
| |
| * add a special API to get a whitelist bitvectorArne Juul2020-05-146-4/+35
| |
* | Merge pull request #13224 from ↵Tor Egge2020-05-1431-45/+544
|\ \ | |/ |/| | | | | vespa-engine/toregge/calculate-transient-memory-usage-for-attribute-vector-load Calculate transient memory usage for load of attribute vector.
| * Update class comments.Tor Egge2020-05-133-4/+7
| |
| * Expose transient memory usage as a metric.Tor Egge2020-05-135-0/+11
| |
| * Add unit test for AttributeUsageSamplerFunctor.Tor Egge2020-05-133-0/+143
| |
| * Consolidate code to calculate transient memory usage for attributeTor Egge2020-05-135-64/+104
| | | | | | | | vector load.
| * Calculate transient memory usage for load of attribute vector.Tor Egge2020-05-1223-18/+320
| |
* | Merge pull request #13234 from ↵Arne H Juul2020-05-141-2/+2
|\ \ | | | | | | | | | | | | vespa-engine/arnej/use-global-filter-in-nn-blueprint Arnej/use global filter in nn blueprint
| * | own the filter in a class and use shared_from_thisArne Juul2020-05-121-2/+2
| |/
* | Merge pull request #13216 from vespa-engine/balder/limit-number-of-lids-per-fileHenning Baldersheim2020-05-132-0/+4
|\ \ | |/ |/| Account and limit number of lids per file to reduce spkies during com…
| * Account and limit number of lids per file to reduce spkies during compaction.Henning Baldersheim2020-05-112-0/+4
| | | | | | | | | | This is to avoid ending up with a very long listof lids that are removed but not compacted away due to lidspace compaction being disabled when removes and delets are too frequent.
* | Merge pull request #13215 from vespa-engine/geirst/modern-cpp-2Henning Baldersheim2020-05-116-31/+32
|\ \ | |/ |/| Use std::make_unique and auto.
| * Use std::make_unique and auto.Geir Storli2020-05-116-31/+32
| |
* | actually call set_global_filter (with empty SP for now)Arne Juul2020-05-111-0/+7
|/
* Merge pull request #13190 from ↵Geir Storli2020-05-093-6/+99
|\ | | | | | | | | vespa-engine/toregge/add-get-transient-memory-usage-to-attribute-initializer Add get_transient_memory_usage() method to AttibuteInitializer.
| * Verify that transient memory is zero when attribute to load doesn't haveTor Egge2020-05-081-2/+10
| | | | | | | | a posting list.
| * Add get_transient_memory_usage() method to AttibuteInitializer.Tor Egge2020-05-083-6/+91
| |
* | Merge pull request #13194 from ↵Henning Baldersheim2020-05-082-5/+16
|\ \ | | | | | | | | | | | | vespa-engine/geirst/improve-lid-space-compaction-log-messages Improve log messages for when job is blocked/un-blocked/disabled/re-e…
| * | Improve log messages for when job is blocked/un-blocked/disabled/re-enabled.Geir Storli2020-05-082-5/+16
| |/
* / Use nested namespaces.Geir Storli2020-05-0828-102/+51
|/
* Update comments.Henning Baldersheim2020-05-051-1/+1
|
* Add async update and followup on PR comments.Henning Baldersheim2020-05-055-41/+37
|
* Implement async putHenning Baldersheim2020-05-042-13/+10
| | | | Implement async remove.
* Rename namespace search::btree to vespalib::btree.Tor Egge2020-05-045-12/+12
| | | | Rename namespace search::datastore to vespalib::datastore.
* Actually propagate result.Henning Baldersheim2020-05-028-31/+25
|
* Remove deprecated BucketIntegrityCheckerTor Brede Vekterli2020-04-304-15/+0
| | | | Not in use after VDS was removed.
* Merge pull request #13113 from vespa-engine/balder/make-put-asyncHenning Baldersheim2020-04-3021-138/+228
|\ | | | | - Add async interface to put
| * Add class comments and resolve PR comments.Henning Baldersheim2020-04-302-6/+17
| |
| * - Add async interface to putHenning Baldersheim2020-04-2921-137/+216
| | | | | | | | | | - Use MessageTracker for keeping context. - implement putAsync, but still use it synchronously.
* | Resource sample interval from 20s => 10sHenning Baldersheim2020-04-301-2/+2
| |
* | If we enter blocking state we should sample every second in order to quickly ↵Henning Baldersheim2020-04-302-5/+14
|/ | | | detect when we are out of it.
* Use forward declaration of OperationListener in header files.Geir Storli2020-04-297-12/+15
|
* Add tracking of remove operations rate and use this to consider blocking lid ↵Geir Storli2020-04-2912-38/+147
| | | | | | | | | space compaction. During a period with a high rate of remove operations, there is no use running lid space compaction as this will interfere with the remove operations, increasing latency of those. Moving a document as part of lid space compaction is a costly operation (similar to putting the document in the first place) and it typically uses both the index and attribute writer thread pools.
* Improve tracking of remove batch rate used to consider to block lid space ↵Geir Storli2020-04-2926-59/+377
| | | | | | | compaction. This is also a preparation for tracking the rate of regular remove operations, and use this to consider to block lid space compaction.
* Remove flush from provider interface.Henning Baldersheim2020-04-272-7/+0
|
* Move BufferWriter back to searchlib.Tor Egge2020-04-252-2/+2
|
* Reduce code duplication. Rand48 class is in vespalib.Tor Egge2020-04-251-13/+13
|