summaryrefslogtreecommitdiffstats
path: root/searchcore
Commit message (Collapse)AuthorAgeFilesLines
* Tag cpu category for tasks run in the proton shared executor.Geir Storli2022-01-289-35/+40
|
* Tag maintenance job tasks with cpu category COMPACT.Geir Storli2022-01-284-13/+18
|
* Tag proton thread pools with CPU categories.Geir Storli2022-01-274-34/+50
|
* Merge pull request #20953 from vespa-engine/geirst/align-resource-usage-metricsHenning Baldersheim2022-01-279-59/+131
|\ | | | | Align resource usage metrics with what is tracked and reported to the…
| * Align resource usage metrics with what is tracked and reported to the ↵Geir Storli2022-01-279-59/+131
| | | | | | | | | | | | | | cluster controller. New disk and memory metrics are added with improved names. The old metrics will be removed on Vespa 8.
* | Merge pull request #20943 from vespa-engine/balder/avoid-explicit-close-1Henning Baldersheim2022-01-263-39/+26
|\ \ | | | | | | - assert result of Close() in destructor of FastOS_File and FastOS_Bu…
| * | - assert result of Close() in destructor of FastOS_File and FastOS_BufferedFile.Henning Baldersheim2022-01-263-39/+26
| |/ | | | | | | | | | | | | - Check result of Close() - Check result of Sync() - Scope FastOS_File to avoid explicit Close(). - NULL -> nullptr
* | Merge pull request #20932 from ↵Geir Storli2022-01-261-2/+2
|\ \ | | | | | | | | | | | | vespa-engine/toregge/adjust-replay-throttling-policy-max-window-size Adjust default replay_throttling_policy.max_window_size in proton config.
| * | Adjust default replay_throttling_policy.min_window_size in proton config.Tor Egge2022-01-261-1/+1
| | |
| * | Adjust default replay_throttling_policy.max_window_size in proton config.Tor Egge2022-01-251-1/+1
| |/
* / tag match/summary engines with READ cpu usageHåvard Pettersen2022-01-262-3/+10
|/ | | | tag llvm compilation tasks with SETUP cpu usage
* Merge pull request #20924 from vespa-engine/toregge/throttle-replayGeir Storli2022-01-259-8/+81
|\ | | | | Throttle replay.
| * Adjust comments.Tor Egge2022-01-252-2/+2
| |
| * Throttle replay.Tor Egge2022-01-259-8/+81
| |
* | Merge pull request #20915 from vespa-engine/havardpe/proton-cpu-util-metricsGeir Storli2022-01-254-1/+36
|\ \ | |/ |/| added cpu util metrics
| * use separate metrics in a metric setHåvard Pettersen2022-01-253-15/+30
| | | | | | | | values are kept unnormalized
| * added cpu util metricsHåvard Pettersen2022-01-244-1/+21
| |
* | Add comment describing why a replay feed token should not be passed to ↵Tor Egge2022-01-241-0/+9
| | | | | | | | GidToLidChangeHandler.
* | Don't pass replay feed token to GidToLidChangeHandler.Tor Egge2022-01-241-1/+2
|/
* Update class comment.Tor Egge2022-01-211-1/+3
|
* Prepare for throttling of replay.Tor Egge2022-01-2116-44/+148
|
* Only require vespalib::Executor instead of vespalib::ThreadExecutorHenning Baldersheim2022-01-1915-45/+29
|
* Merge pull request #20861 from vespa-engine/arnej/special-document-only-fieldsetArne H Juul2022-01-191-0/+9
|\ | | | | add special '[document]' fieldset
| * add special '[document]' fieldsetArne H Juul2022-01-181-0/+9
| | | | | | | | | | | | | | * specially handled like [all] and [docid] * depends on the document type having a configured fieldset named '[document]' * should make it possible to run vespa-visit -l '[document]' and get only the fields declared inside the document for any and all document types.
* | Wire in control of whether taskLimit is hard.Henning Baldersheim2022-01-185-10/+24
|/
* Remove sync apply bucket diff.Tor Egge2022-01-184-14/+1
|
* Speed up persistence conformance test.Tor Egge2022-01-151-1/+7
|
* Subtract transient resource usage from the absolute usage before reporting ↵Geir Storli2022-01-142-4/+19
| | | | | | | to the cluster controller. This ensures that the transient resource usage is covered by the resource headroom on the content node, instead of leading to feed blocked due to natural fluctuations.
* Merge pull request #20782 from ↵Geir Storli2022-01-131-16/+16
|\ | | | | | | | | vespa-engine/toregge/calculate-transient-disk-space-due-to-fusion Consider disk space used by fusion indexes beyond current fusion index or
| * Rename ActiveDiskIndexes to DiskIndexes.Tor Egge2022-01-131-16/+16
| | | | | | | | Rename IndexDiskDirActiveState to IndexDiskDirState.
| * Consider disk space used by fusion indexes beyond current fusion index orTor Egge2022-01-121-2/+2
| | | | | | | | indexes before current fusion index as transient disk space.
* | Set defaults in config defs and ModelContext api to improve bucket merge ↵Geir Storli2022-01-122-4/+5
|/ | | | | | performance on content nodes. These are the same defaults set for the feature flags in https://github.com/vespa-engine/vespa/pull/20759.
* Reduce default summary cache memory from 5% to 4% all over.Henning Baldersheim2022-01-111-1/+1
|
* Merge pull request #20740 from ↵Geir Storli2022-01-1012-55/+80
|\ | | | | | | | | vespa-engine/geirst/prepare-sampling-of-current-transient-disk-usage Prepare for tracking the current transient disk usage across all docu…
| * Prepare for tracking the current transient disk usage across all document dbs.Geir Storli2022-01-1012-55/+80
| | | | | | | | | | The next step will be sampling the total disk usage of all current fusion indexes and report this as the transient disk usage.
* | Merge pull request #20729 from ↵Harald Musum2022-01-102-5/+6
|\ \ | | | | | | | | | | | | vespa-engine/revert-20728-revert-20724-balder/default-disk-bloat-at-25-percent Revert "Revert "Balder/default disk bloat at 25 percent""
| * | Update test after default disk bloat changed from 0.20 to 0.25Henning Baldersheim2022-01-101-3/+4
| | |
| * | Revert "Revert "Balder/default disk bloat at 25 percent""Henning Baldersheim2022-01-101-2/+2
| | |
| * | Revert "Balder/default disk bloat at 25 percent"Harald Musum2022-01-101-2/+2
| | |
| * | Default disk bloat changed from 20% to 25%.Henning Baldersheim2022-01-101-2/+2
| |/
* / Simplify use of search::diskindex::Fusion class.Tor Egge2022-01-092-33/+34
|/
* Sample and track the current transient memory usage across all document dbs.Geir Storli2022-01-0729-243/+113
| | | | | | | | | | | Previously, the transient memory usage was the theoretical worst case needed during loading of attributes. Now, we change this to being a snapshot of the total memory used by all memory indexes. In a simular fashion we plan to track the transient disk usage of all ongoing fusion disk indexes. Eventually, the transient resource usage will be subtracted from the absolute resource usage, before this is reported to the cluster controller in order to decide whether to block feeding. By doing this the transient resource usage is covered by the headroom on the content node, instead of leading to feed blocked due to natural fluctuations.
* Rewrite unit test to use GTest.Geir Storli2022-01-072-53/+54
|
* Remove unused functions.Geir Storli2022-01-071-2/+0
|
* Revert "Revert "Balder/refactor docentry""Henning Baldersheim2022-01-072-30/+31
|
* Revert "Balder/refactor docentry"Arnstein Ressem2022-01-072-31/+30
|
* - Flags -> Enum.Henning Baldersheim2022-01-062-12/+12
| | | | - Consistently use DocEntryList as type for std::vector<spi::DocEntry::UP>
* Only care about size of payload. Also add payload containing only doctype ↵Henning Baldersheim2022-01-061-5/+5
| | | | and gid
* Use enum class for the flags.Henning Baldersheim2022-01-062-13/+14
|
* Simplify by avoid both DocumentSize and PersistedDocumentSize. That is the same.Henning Baldersheim2022-01-062-20/+16
|