summaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Revert "relative estimate"Henning Baldersheim2023-12-092-23/+4
|
* relative estimateHåvard Pettersen2023-12-082-4/+23
|
* revert default setting for use_estimate_for_fetch_postings to false and add ↵Henning Baldersheim2023-12-061-1/+2
| | | | TODO.
* Wire in use-estimate-for-fetch-postings into ExecuteInfo.Henning Baldersheim2023-12-064-7/+10
|
* Merge pull request #29551 from ↵Henning Baldersheim2023-12-0517-56/+20
|\ | | | | | | | | vespa-engine/balder/gc-use-shared-executor-for-warmup Use shared executor for warmup and GC warmup executor.
| * Use shared executor for warmup and GC warmup executor.Henning Baldersheim2023-12-0517-56/+20
| |
* | Consistent namingHenning Baldersheim2023-12-051-1/+1
| |
* | - Control creation of temporary postinglists during fetchPostings for ↵Henning Baldersheim2023-12-045-9/+13
|/ | | | non-strict iterators.
* Avoid timeout during grouping leaving distributionKey unset. Populate it ↵Henning Baldersheim2023-11-304-26/+15
| | | | right after completing grouping.
* Add InTerm to backend.Tor Egge2023-11-248-0/+10
|
* Avoid casting by adding asXXX methods to the Blueprint interface.Henning Baldersheim2023-11-223-14/+12
|
* Avoid dereferencing first item in an empty vector.Henning Baldersheim2023-11-201-1/+1
|
* - We are now always nesting multivalue grouping for indexed search.Henning Baldersheim2023-11-205-40/+31
|
* Add flag for marking phrase always expensive.Henning Baldersheim2023-11-196-8/+57
|
* Merge pull request #29352 from vespa-engine/balder/create-iterator-outside-guardHenning Baldersheim2023-11-172-8/+17
|\ | | | | Do not hold the guard when creating the iterator.
| * Use structured captureHenning Baldersheim2023-11-171-2/+2
| |
| * Use a pair as suggested in code reviewHenning Baldersheim2023-11-172-5/+6
| |
| * Do not hold the guard when creating the iterator.Henning Baldersheim2023-11-162-8/+16
| |
* | Merge pull request #29369 from vespa-engine/balder/gc-unused-split-parameterHenning Baldersheim2023-11-176-60/+19
|\ \ | | | | | | Fully GC unused parameter as we now always split phrases.
| * | Fully GC unused parameter as we now always split phrases.Henning Baldersheim2023-11-176-60/+19
| |/
* | Merge pull request #29355 from vespa-engine/balder/avoid-limiting-not-necessaryHenning Baldersheim2023-11-172-2/+12
|\ \ | | | | | | Balder/avoid limiting not necessary
| * | If hit_rate is below 1% drop match phase limiting. It has too high fixed ↵Henning Baldersheim2023-11-162-2/+12
| |/ | | | | | | cost and will liklely make things worse.
* / Keep item type and features separate.Tor Egge2023-11-171-1/+1
|/
* fix tsan errorHåvard Pettersen2023-11-132-3/+3
| | | | reported in VESPA-27749
* More detailed posting store memory usage in attribute explorer.Geir Storli2023-11-101-2/+7
|
* Merge pull request #29284 from ↵Tor Brede Vekterli2023-11-094-21/+62
|\ | | | | | | | | vespa-engine/vekterli/include-doctype-and-gid-with-metadata-doc-entries Include doc type name and GID in metadata iteration results
| * Simplify by passing in and storing the `DocTypeName` verbatimTor Brede Vekterli2023-11-084-54/+15
| |
| * Include doc type name and GID in metadata iteration resultsTor Brede Vekterli2023-11-084-20/+100
| | | | | | | | | | | | | | | | | | | | | | | | | | | | Document type is fetched from the associated `IPersistenceHandler` on-demand; it is assumed the lifetime of the pointer must be valid for the entire lifetime of the iterator itself, as the latter holds a valid handler snapshot. For simplicity, it's possible to _not_ pass in a handler, in which case the doc type name will be implicitly empty. Some expected `DocEntry` sizes have been adjusted, as we now report the size of the document type and GID alongside the base type size.
* | Sameelement behaves like an and with extra constraints.Henning Baldersheim2023-11-074-9/+9
|/ | | | So it should behave the sameway during fetchPostings too.
* Merge pull request #29270 from ↵Henning Baldersheim2023-11-071-1/+3
|\ | | | | | | | | vespa-engine/balder/do-not-offer-post-filters-further-down Balder/do not offer post filters further down
| * Only offer filter for insertion further down if it is a prefilterHenning Baldersheim2023-11-071-1/+3
| |
* | Merge pull request #29269 from ↵Geir Storli2023-11-0717-78/+110
|\ \ | |/ |/| | | | | vespa-engine/geirst/control-resource-usage-when-in-maintenance Control resource usage when node in maintenance
| * Also tune or turn off background jobs when content node is in maintenance.Geir Storli2023-11-0716-50/+91
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | Previously the following has been adjusted when the node is retired: 1) Lid space compaction - turned off. 2) Flush engine strategy - tuned to reduce disk and CPU usage. 3) Attribute vector compaction - tuned to reduce memory allocations and CPU usage. In a node retirement scenario documents are being removed from the node, and eventually the node is deleted. Without the adjustments above a lot of resources are spent "fixing" the results of removing documents, and the process just takes a lot longer. A similar set of challenges can occur when a node is set in maintenance, especially if the node transitions from retired to maintenance. E.g. this happens when the Vespa version is upgraded in Vespa Cloud. With this change the resource usage of background jobs are kept in check for both a retired node and a node in maintenance.
| * Rewrite to use GTest.Geir Storli2023-11-072-35/+26
| |
* | Merge pull request #29266 from ↵Henning Baldersheim2023-11-076-6/+25
|\ \ | | | | | | | | | | | | vespa-engine/vekterli/expose-doc-type-name-from-persistence-handler Expose document type name from `IPersistenceHandler` interface
| * | Expose document type name from `IPersistenceHandler` interfaceTor Brede Vekterli2023-11-076-6/+25
| |/
* | Merge pull request #29264 from ↵Henning Baldersheim2023-11-072-1/+4
|\ \ | | | | | | | | | | | | vespa-engine/toregge/extend-persistence-conformance-test-wrt-remove-by-gid Test remove by gid for nonexisting gid and for gid with tombstone.
| * | Test remove by gid for nonexisting gid and for gid with tombstone.Tor Egge2023-11-072-1/+4
| |/
* | If strict you will be earlier, and no assumptions on hit rate can be taken.Henning Baldersheim2023-11-071-1/+1
| |
* | If match-phase limiting has concluded that a post filter is most efficient,Henning Baldersheim2023-11-075-18/+20
|/ | | | we must only generate posting lists if it is actually benefiscal. If not the fixed cost is too high.
* Add removeByGidAsync() to spi.Tor Egge2023-11-067-0/+45
|
* Not possible to do safe cast here. Must leav as is and wait until we clean ↵Henning Baldersheim2023-11-061-15/+9
| | | | up with proper get interfaces.
* Make test for target numdocs auto-tuning aware of sanitizersTor Brede Vekterli2023-11-061-2/+14
| | | | | Test must mirror the logic used for enabling auto-tuning, or it will fail when testing on a sanitizer-instrumented build.
* Not logging anything hereHenning Baldersheim2023-11-041-3/+0
|
* Use explicit singlevalue interface.Henning Baldersheim2023-11-041-12/+18
|
* Revert "No need to specify your own namespace."Henning Baldersheim2023-11-041-18/+12
|
* - Add maxHits param that can be used to compute a realistic(low) hit rate ↵Henning Baldersheim2023-11-038-20/+25
| | | | when fetching summaries (summary features).
* When fetching summaries we do not require a strict search iterator.Henning Baldersheim2023-11-033-4/+5
| | | | We are iterating the document ids explicit anyway.
* Use explicit singlevalue interface.Henning Baldersheim2023-11-031-12/+18
|
* Do not preallocate doc store based on node memory if sanitizers are enabledTor Brede Vekterli2023-11-021-1/+14
| | | | | | | If sanitizers are enabled, `mmap`-allocations may be intercepted and allocated pages may be implicitly touched+committed. This tends to explode when testing locally, so fall back to configured initial num-docs if this is the case.