summaryrefslogtreecommitdiffstats
path: root/config-model
Commit message (Collapse)AuthorAgeFilesLines
* Revert "Bratseth/tensor type info in documents"Jon Bratseth2017-01-1267-892/+326
|
* Merge pull request #1495 from yahoo/bratseth/tensor-type-info-in-documentsJon Bratseth2017-01-1267-326/+892
|\ | | | | Bratseth/tensor type info in documents
| * Merge branch 'master' into bratseth/tensor-type-info-in-documentsJon Bratseth2017-01-121-1/+0
| |\
| * | Tensor type information in tensor document fieldsJon Bratseth2017-01-1265-282/+867
| | |
| * | Maintain TensorType in documentsJon Bratseth2017-01-104-56/+37
| | |
* | | Improve name.Haakon Dybdahl2017-01-121-1/+1
| | |
* | | Update with new overrides.Haakon Dybdahl2017-01-121-1/+2
| |/ |/| | | | | | | | | | | Removing old staging override, not used, was intended for old screwdriver config. Adding two new overrides. One is for enabling the tests while it is default off (in the beginnging) and one for disabling it after it has been enabled as default. The enable override should be deprecated after it has been turned on as default.
* | Remove support for 'chain' under 'http'.gjoranv2017-01-111-1/+0
|/ | | | - No longer allowed in rnc schema.
* Merge pull request #1354 from yahoo/kraune/disklimit-0.8Geir Storli2017-01-021-1/+1
|\ | | | | set new disk feed block at 80%
| * set new disk feed block at 80%Kristian Aune2016-12-191-1/+1
| | | | | | | | - VESPA-3494
* | Merge pull request #1381 from yahoo/bratseth/indexed-tensor-functionsHenning Baldersheim2016-12-223-1/+47
|\ \ | | | | | | Bratseth/indexed tensor functions
| * | Add tensor generate functionsJon Bratseth2016-12-223-1/+47
| | |
* | | Remove methods and constants no longer in useHarald Musum2016-12-221-9/+0
| | | | | | | | | | | | * Move hosted vespa tenant name
* | | Revert "Remove methods and constants no longer in use"Harald Musum2016-12-221-0/+9
| | |
* | | Remove methods and constants no longer in useHarald Musum2016-12-221-9/+0
| | |
* | | Expose missing distributor visitor metrics in modelTor Brede Vekterli2016-12-201-3/+4
|/ /
* / Track API changesJon Bratseth2016-12-203-6/+6
|/
* Expose detailed memory usage metrics from proton.Geir Storli2016-12-161-0/+28
|
* Merge pull request #1309 from yahoo/hmusum/minor-cleanupValerij Fredriksen2016-12-131-2/+4
|\ | | | | Use explicit imports and remove unused argument
| * Use explicit imports and remove unused argumentHarald Musum2016-12-091-2/+4
| |
* | Fix unit tests after update of memory resource limitsHarald Musum2016-12-121-2/+2
| | | | | | | | * Fixes PR-1304 and reverts PR-1306
* | Update expected result.Henning Baldersheim2016-12-121-1/+1
| |
* | Store flavor for nodes provisioned by NodeRepositoryProvisioner.gjoranv2016-12-093-6/+18
|/ | | | | - in config-provisioning:HostSpec and config-model:HostResource. - will be used by metrics-proxy, but should be useful overall.
* Revert "Always use 16 distribution bits on hosted systems"Tor Brede Vekterli2016-12-062-7/+5
|
* Allow configuring docker registry and Docker Vespa base imageHarald Musum2016-12-053-0/+14
|
* Correct over-eager renamingJon Bratseth2016-12-0414-23/+23
|
* Towards always typed, fully specified tensorsJon Bratseth2016-12-041-4/+4
| | | | | | | | | - Tensor addresses do not repeat dimensions. - Tensor addresses must provide a value for all dimensions. - Tensor dimensions are not serialized in JSON (but still are in binary). - Tensor types are required everywhere, except a workaround for JSON deserialization. - Tensor operations are about 50% faster. - Tensor join of two tensors in the same space is about 4000% faster.
* Rename methodJon Bratseth2016-12-0314-23/+23
|
* Always use 16 distribution bits on hosted systemsJon Bratseth2016-12-022-5/+7
| | | | | This avoids distribution skew with few bits and massive redistribution events when nodes are deactivated after reducing the number of nodes.
* Do not use null as arguments when creating content clusterHarald Musum2016-12-021-38/+40
|
* Remove unused variable and use root instead of creating new oneHarald Musum2016-12-022-2/+1
|
* Do not use null as arguments in testsHarald Musum2016-12-024-28/+28
|
* Merge branch 'master' into arnej/count-memory-mappingsArne H Juul2016-11-302-34/+10
|\ | | | | | | | | Conflicts: searchcore/src/vespa/searchcore/proton/metrics/resource_usage_metrics.cpp
| * Propagate set/getChildrenJon Bratseth2016-11-282-34/+10
| |
* | rename metrics, add open file descriptorsArne H Juul2016-11-291-2/+4
| | | | | | | | | | | | * renamed count metric to just "memory_mappings" * count open filedescriptors in container * add "jdisc." prefix to new jdisc metrics
* | added in wrong place, now with correct nameArne H Juul2016-11-251-1/+1
| |
* | add memory_mappings_count to vespa metric setArne H Juul2016-11-251-0/+2
|/
* Fix Perl syntax and use printfHarald Musum2016-11-251-1/+1
|
* Add deprecation warningsHarald Musum2016-11-251-0/+2
|
* Remove special handling of deployment of routing applicationHarald Musum2016-11-232-139/+5
| | | | | | * Routing application uses nodes from node repo in all zones now, handle as other applications VESPA-5118
* Set system in config server configHarald Musum2016-11-222-3/+4
|
* Add validation id for staging tests.Haakon Dybdahl2016-11-211-1/+2
|
* Merge pull request #1133 from yahoo/bratseth/ranking-expression-parsing-speedHenning Baldersheim2016-11-1823-160/+2991
|\ | | | | Bratseth/ranking expression parsing speed
| * Store rank properties in compressed formJon Bratseth2016-11-181-15/+28
| | | | | | | | | | | | We run out of memory on mutlitenant config servers due to spending too mcuh memory storing large rank profiles. This removes most of that cost by storing the rank profile as compressed bytes.
| * Prepare for compression of rank profilesJon Bratseth2016-11-1823-158/+2976
| |
* | Revert "Disable validation of deployment.xml"Harald Musum2016-11-171-2/+0
| |
* | region->systemJon Marius Venstad2016-11-161-1/+1
| |
* | Make sure config server picks up correct system nameHarald Musum2016-11-161-1/+3
|/
* Make sure that 'system' yinst setting is propagated to configserver configHarald Musum2016-11-143-2/+8
|
* I assume the intention was to first run clean, then make all.Henning Baldersheim2016-11-111-1/+1
| | | | Specifying both will make then race each other when building with many threads.