summaryrefslogtreecommitdiffstats
path: root/node-repository
Commit message (Collapse)AuthorAgeFilesLines
* Do not try to deploy after operator changes on all config serversHarald Musum2019-08-141-5/+14
| | | | | | If application is not deployed locally on a config server, the last deployment time used will be epoch. Change this so that deployment will only be attempted on the server where the application has been deployed locally.
* Ress resources is not compatible with more resourcesValerij Fredriksen2019-08-131-3/+3
|
* Allocate nodes through the host for tests in devValerij Fredriksen2019-08-134-33/+37
|
* Fix off-by-one error in testerValerij Fredriksen2019-08-132-2/+2
|
* Always prepare with Capacity rather than countValerij Fredriksen2019-08-132-13/+10
|
* flavor -> nodeResourcesValerij Fredriksen2019-08-131-10/+10
|
* Use flag for node resource memory for admin nodesHarald Musum2019-08-123-4/+8
|
* Add throttled host metricValerij Fredriksen2019-08-072-6/+13
|
* Nonfunctional changes onlyJon Bratseth2019-08-051-4/+4
|
* Remove code that handles node resources based on flavors for cdHarald Musum2019-08-021-15/+8
| | | | | We don't use flavors in cd tests any more, only node resources, remove unused code
* Return 409 with error code LOAD_BALANCER_NOT_READY if deployment fails due ↵Harald Musum2019-08-012-18/+1
| | | | | | | | | to that Move LoadBalancerServiceException to config-provisioning so that it can be used from config server as well. Return 409 and LOAD_BALANCER_NOT_READY as error code if we get a LoadBalancerServiceException. Retry deploy on LOAD_BALANCER_NOT_READY.
* Revert "Return 409 with error code TRANSIENT_ERROR when getting ↵Harald Musum2019-08-0112-11/+28
| | | | TransientExce…"
* Move LoadBalancerServiceException to config-provisioning moduleHarald Musum2019-08-012-18/+1
|
* Move some exceptions to its own package (making them not part of public API)Harald Musum2019-08-0111-11/+11
|
* Set memory for admin cluster node to 2 Gb in CDHarald Musum2019-07-311-1/+6
|
* Use 2.5 Gb memory for nodes in admin cluster in CDHarald Musum2019-07-291-1/+1
| | | | | | We have reduced max heap sizes for logserver-container and metricsproxy-container since we started using 3 Gb, so it should be possible to reduce memory for these nodes, trying with .5 Gb less
* Remove noop code (does not return)Harald Musum2019-07-261-3/+0
| | | | | Since we haven't noticed any issues with this not working, remove to make code simpler
* Remove unused fields from API responseMartin Polden2019-07-193-13/+3
|
* Made the capacity endpoint accept a comma-separated list instead of JSON.mgimle2019-07-183-17/+8
| | | | Made the endpoint accept calls which do not end with a trailing slash.
* Implemented suggested changes to CapacityChecker.mgimle2019-07-153-60/+61
|
* Added endpoint for explaining the conclusions of the host capacity checker.mgimle2019-07-156-0/+273
|
* Separated out the capacity checking part of the capacity report maintainer.mgimle2019-07-155-429/+568
| | | | | Made aws zones excluded from the capacity calculation because the metric is nonsensical for an automatically scaling zone. Fixed some missing renames.
* Removed Agent enums pertaining to the now-removed NodeRetirermgimle2019-07-122-3/+1
|
* Replace 'jdisc' with 'container' in node-repositorygjoranv2019-07-112-4/+4
|
* Reduced size of node repository dump.mgimle2019-07-091-1/+1
|
* Renamed from NodeAlerter to CapacityReportMaintainer.mgimle2019-07-095-34/+33
|
* Incorporated suggested changes to the NodeAlerter.mgimle2019-07-081-104/+109
|
* Added more comments, made private members private,mgimle2019-07-083-107/+171
| | | | | | | | made static inner class candidates static, moved filtering of uninteresting hosts and tenants to the alerter from the test case generator, added a test for overcommitted nodes, logically segmented the generated test cases, removed explicit null checks in favour of Optionals, changed some inner class initialization behaviour.
* Implemented the Node Alerter maintainer and metric.mgimle2019-07-086-0/+848
| | | | | | | | | | | | | | | This metric attempts to determine how many hosts it's possible to lose before there's no place to fit its tenants, by finding a "shortest path to failure". Since finding the actual path is np-hard, this maintainer constructs a heuristic based on "repeated removals", and finds greedily finds a path to failure with it. The Node Alerter also exposes the "overcommittedNodes" metric, counting how many hosts have children expecting more resources than it can provide. Finally, this commit adds an obfuscated dump of data from zookeeper, useful for running tests which require a node repository which reflects reality.
* Update javadocMartin Polden2019-07-031-3/+1
|
* Merge pull request #9931 from ↵Martin Polden2019-07-012-2/+2
|\ | | | | | | | | vespa-engine/freva/move-serializer-to-non-public-api Move serializers to non public api
| * Move NetworkPortsSerializer to non-public-api serialization packageValerij Fredriksen2019-07-012-2/+2
| |
* | Reduced log for transient errors in ApplicationMaintainerValerij Fredriksen2019-06-301-0/+4
| |
* | Simplify OperatorChangeApplicationMaintainerValerij Fredriksen2019-06-302-23/+17
| |
* | Simplify PeriodicApplicationMaintainerValerij Fredriksen2019-06-302-16/+7
| |
* | Always activate through MockDeployerValerij Fredriksen2019-06-303-76/+51
| |
* | Ignore TransientException in NodeFailer and RetiredExpirerValerij Fredriksen2019-06-292-2/+14
|/
* Fix TODOMartin Polden2019-06-271-2/+0
|
* Read and write all load balancers on constructionMartin Polden2019-06-272-7/+19
|
* Avoid reading all load balancers on prepareMartin Polden2019-06-272-9/+10
|
* Remove unused rotations parameterMartin Polden2019-06-2620-72/+56
|
* Remove support for legacy serialization formatMartin Polden2019-06-253-48/+5
| | | | New format is now used everywhere.
* Revert "Do not require changeAt to be after previous value"Martin Polden2019-06-251-0/+4
| | | | This reverts commit a1bd5a9f0e9a72f24b96e08755deb6c17e0e7b41.
* Force load balancer reconfig when in non-active stateMartin Polden2019-06-256-13/+81
|
* Fix exception messageMartin Polden2019-06-251-1/+1
|
* Log realsMartin Polden2019-06-251-2/+2
|
* Rename testMartin Polden2019-06-241-1/+1
|
* Never provision a new LB in activateMartin Polden2019-06-242-4/+29
|
* Fix nameMartin Polden2019-06-241-3/+3
|
* Use common exception class for transient provisioning exceptionsMartin Polden2019-06-212-1/+25
|