aboutsummaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* add more information to exception messageArne Juul2021-08-301-1/+2
|
* signal timeout instead of continuing triggering core dumpsArne Juul2021-08-301-0/+4
|
* Merge pull request #18859 from ↵Geir Storli2021-08-252-0/+11
|\ | | | | | | | | vespa-engine/toregge/get-paged-setting-from-live-attribute-config Get paged setting from live attribute config.
| * Get paged setting from live attribute config.Tor Egge2021-08-252-0/+11
| |
* | info -> debugHenning Baldersheim2021-08-251-1/+1
| |
* | Add some debug loggingHenning Baldersheim2021-08-251-1/+1
| |
* | Add more information to error messageHenning Baldersheim2021-08-251-1/+1
|/
* Report max attribute address space usage in state explorer.Geir Storli2021-08-235-12/+34
|
* Expose address space usage for all attribute components in explorer.Geir Storli2021-08-201-2/+3
|
* Report max address space used in attribute vector components from content ↵Geir Storli2021-08-203-47/+39
| | | | | | | nodes (proton) to the cluster controller. This is more generic than explicit address space values for enum store and multi value. This is used in the cluster controller to determine whether to block external feed.
* Use max address space used in attribute vector components to determine ↵Geir Storli2021-08-2010-111/+45
| | | | whether to block put and update operations in proton.
* Add metric for the max address space used among components in all attribute ↵Geir Storli2021-08-203-4/+9
| | | | vectors in a document db.
* Track max address space usage among components in attributes vectors in all ↵Geir Storli2021-08-207-9/+86
| | | | sub databases.
* Also provide used and on-hold memory in overview.Henning Baldersheim2021-08-151-0/+2
|
* Wire in the shared executor for use when loading attributes.Henning Baldersheim2021-08-134-11/+18
|
* Add an executor to the AttributeVector::load/onLoad interface so attributes ↵Henning Baldersheim2021-08-122-2/+2
| | | | can use multithread load if feasible.
* Add some debug logging.Henning Baldersheim2021-08-093-2/+15
|
* Split current global_filter_limit into global_filter.lower_limit/upper_limit.Henning Baldersheim2021-08-043-12/+18
| | | | | | | | | If estimated_hits < lower_limit no filter is set which will cause fallback to bruteforce. If estimated_hits in [lower_limit, upper_limit] apply global filter. if estimated_hits > upper_limit an empty filter is set. This will avoid the filter setup cost. So if the filter has a huge setup cost, you can reduce upper_limit to a number below 1.0 and instead increase target_num_hits similarly. Setting target_num_hits to 1.0/upper_limit * 1.2 should give similar recall. This will add a 20% safety to handle correlation of filter and NearestNeightbor calculation.
* Also perform topk when you do not use a global filter.Henning Baldersheim2021-08-031-0/+4
|
* Generalize how address space usage is reported from attribute vectors.Geir Storli2021-07-153-14/+18
| | | | | This prepares for reporting address space usage for more attribute vector components, not only enum store and multi-value mapping.
* Fix typo.Geir Storli2021-07-152-3/+3
|
* Optimize assign updates to tensor attribute with hnsw index by handling them ↵Geir Storli2021-07-143-25/+128
| | | | | | as two phase put operations. This ensures that the costly part of inserting into the hnsw index is done in parallel in the prepare put step.
* Eliminate race in flush engine unit test which caused premature wakeup.Tor Egge2021-07-091-1/+8
|
* Adjust reporting of disk and memory usage in proton.Tor Egge2021-07-0710-9/+135
|
* Omit filling summary features if specified for the document summary class.Geir Storli2021-07-021-0/+2
|
* GC unused configHenning Baldersheim2021-07-011-11/+0
|
* Revert "Revert "Add support for mallinfo2 with glibc 2.33""Henning Baldersheim2021-06-301-4/+4
|
* Revert "Revert "Revert "Add support for mallinfo2 with glibc 2.33"""Harald Musum2021-06-291-4/+4
|
* Revert "Revert "Add support for mallinfo2 with glibc 2.33""Henning Baldersheim2021-06-291-4/+4
|
* Revert "Add support for mallinfo2 with glibc 2.33"Harald Musum2021-06-291-4/+4
|
* Add support for mallinfo2 with glibc 2.33Henning Baldersheim2021-06-291-4/+4
|
* mallinfo() is only available on linux systems.Tor Egge2021-06-291-0/+6
|
* mallinfo() is deprecated in glibc 2.33.Tor Egge2021-06-291-0/+5
|
* Use the 1_Mi literal for readability.Henning Baldersheim2021-06-281-1/+1
|
* Use mallinfo to report vespamalloc heap size.Henning Baldersheim2021-06-283-2/+9
|
* GC unused members from SearchReplyHenning Baldersheim2021-06-285-9/+1
|
* GC unused 'partition' id.Henning Baldersheim2021-06-261-3/+0
|
* GC unused 'partitionid' and various other members of MonitorReply that are ↵Henning Baldersheim2021-06-262-30/+28
| | | | long gone.
* Various cleanup after reading indexing threading codeHenning Baldersheim2021-06-241-5/+2
|
* Merge pull request #18387 from vespa-engine/balder/gc-dual-semicolonGeir Storli2021-06-241-7/+7
|\ | | | | GC dual semicolon
| * GC dual semicolonHenning Baldersheim2021-06-231-7/+7
| |
* | Merge pull request #18384 from ↵Henning Baldersheim2021-06-231-1/+12
|\ \ | |/ |/| | | | | vespa-engine/geirst/vespa-feed-bm-skip-communicationmanager-thread Add option to skip communicationmanager thread.
| * Add option to skip communicationmanager thread.Geir Storli2021-06-231-1/+12
| | | | | | | | This is originally used when handling messages over mbus.
* | Merge pull request #18386 from vespa-engine/havardpe/onnx-dry-run-implHenning Baldersheim2021-06-233-2/+26
|\ \ | | | | | | dry run onnx models on setup
| * | dry run onnx models on setupHåvard Pettersen2021-06-233-2/+26
| |/
* | Use equality operators.Henning Baldersheim2021-06-233-14/+36
| |
* | Only log when config actually changes.Henning Baldersheim2021-06-233-7/+47
|/
* Merge pull request #18364 from vespa-engine/havardpe/onnx-model-dry-runHenning Baldersheim2021-06-221-0/+1
|\ | | | | add config to perform model dry-run on setup
| * add config to perform model dry-run on setupHåvard Pettersen2021-06-221-0/+1
| |
* | Increase reported replay operation cost for tensor attributes with HNSW index.Tor Egge2021-06-221-1/+1
|/