Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Retire nodes on overcommited hosts | Valerij Fredriksen | 2020-09-11 | 1 | -18/+35 |
| | |||||
* | Make final and pass in NodeList instead of NodeRepository | Valerij Fredriksen | 2020-09-11 | 1 | -1/+1 |
| | |||||
* | Revert "Revert "Bratseth/allocation improvements" MERGEOK" | Jon Bratseth | 2020-09-09 | 4 | -5/+63 |
| | |||||
* | Revert "Bratseth/allocation improvements" | Jon Marius Venstad | 2020-09-09 | 4 | -63/+5 |
| | |||||
* | Don't change node allocation when we cannot fail | Jon Bratseth | 2020-09-09 | 2 | -13/+37 |
| | |||||
* | Reuse inactive nodes also when they are retired first | Jon Bratseth | 2020-09-09 | 3 | -2/+36 |
| | |||||
* | Merge pull request #14321 from vespa-engine/bratseth/autoscale-reaction-time | Valerij Fredriksen | 2020-09-08 | 1 | -1/+1 |
|\ | | | | | Cluster size independent autoscaling reaction time | ||||
| * | Cluster size independent autoscaling reaction time | Jon Bratseth | 2020-09-08 | 1 | -1/+1 |
| | | | | | | | | This makes the window considered about 1 hour regardless of cluster size. | ||||
* | | Avoid renaming DockerImage which is used in the config model | Håkon Hallingstad | 2020-09-08 | 15 | -38/+38 |
| | | |||||
* | | Move to container engine neutral names | Håkon Hallingstad | 2020-09-07 | 15 | -38/+38 |
|/ | |||||
* | Merge pull request #14192 from vespa-engine/mpolden/never-downgrade | Jon Bratseth | 2020-08-28 | 1 | -6/+16 |
|\ | | | | | Never attempt to downgrade OS | ||||
| * | Never attempt to downgrade OS | Martin Polden | 2020-08-28 | 1 | -6/+16 |
| | | |||||
* | | Enforce a minimum vcpu | Jon Bratseth | 2020-08-27 | 1 | -2/+17 |
|/ | | | | | With too little cpu allocated, nodes become unable to perform their duties as citizens of the cloud, such as responding timely to health probes. | ||||
* | Preserve allocation when possible | Jon Bratseth | 2020-08-20 | 3 | -1/+190 |
| | | | | | | | | | | We want to avoid unnecessary small adjustments to capacity. Since the worst case real resources we'll get is generally worse than the actual allocation, we cannot guarantee that we compute the same allocation to request as the one leading to the current allocation. So instead of relying on that we explicitly reuse the current (requested) resources when they are still legal instead of recomputing each time. | ||||
* | Revert "Upgrade to Curator 4" | Harald Musum | 2020-08-17 | 1 | -1/+0 |
| | |||||
* | Upgrade to Curator 4 | Harald Musum | 2020-08-16 | 1 | -0/+1 |
| | |||||
* | Stick to junit for simple test. | Henning Baldersheim | 2020-08-11 | 1 | -3/+1 |
| | |||||
* | Merge pull request #14018 from vespa-engine/freva/fix-retirement-loop | Valerij Fredriksen | 2020-08-11 | 4 | -14/+32 |
|\ | | | | | [VESPA-18652] Fix retirement loop | ||||
| * | Prioritize retiring nodes on spare hosts | Valerij Fredriksen | 2020-08-10 | 1 | -1/+24 |
| | | |||||
| * | Use node type rather than flavor environment to determine node type | Valerij Fredriksen | 2020-08-10 | 3 | -13/+8 |
| | | |||||
* | | Fix recursive wantToRetire & wantToDeprovision patch | Valerij Fredriksen | 2020-08-10 | 1 | -2/+13 |
|/ | |||||
* | Avoid unnecessary fixation of manual clock | Håkon Hallingstad | 2020-07-30 | 1 | -2/+1 |
| | |||||
* | Fix flaky NodeMetricsDbTest | Håkon Hallingstad | 2020-07-30 | 1 | -1/+5 |
| | | | | | | | | | | | ManualClock is initialized internally to Instant.now(). If this instant has a zero subsecond part db.getWindow(clock.instant().minus(Duration.ofHours(30)), Resource.cpu, List.of("host0")).measurementCount() returns 30, while otherwise returns 29. It's hard to say what is correct when falling exactly on this boundary, so this PR avoids the whole problem by ensuring the boundary is not hit (by adding an arbitrary 1 minute). | ||||
* | Take config server lock in node-repository | Martin Polden | 2020-07-22 | 1 | -1/+1 |
| | |||||
* | Inline unnecessary methods | Martin Polden | 2020-07-21 | 4 | -10/+10 |
| | |||||
* | Emit QoS metric for all maintainers | Martin Polden | 2020-07-16 | 10 | -18/+23 |
| | |||||
* | Control maintenance jobs with feature flag | Martin Polden | 2020-07-09 | 14 | -24/+20 |
| | |||||
* | Remove configserver-provision-lb flag | Martin Polden | 2020-06-24 | 1 | -1/+0 |
| | |||||
* | Merge pull request #13679 from vespa-engine/bratseth/prefer-larger-hosts | Jon Bratseth | 2020-06-24 | 1 | -1/+24 |
|\ | | | | | Prefer hosts at least twice the size of the node | ||||
| * | Prefer hosts at least twice the size of the node | Jon Bratseth | 2020-06-24 | 1 | -1/+24 |
| | | |||||
* | | Revert "Revert "From 0 to 1 spare"" | Jon Bratseth | 2020-06-23 | 13 | -19/+57 |
|/ | |||||
* | Revert "From 0 to 1 spare" | Håkon Hallingstad | 2020-06-23 | 13 | -57/+19 |
| | |||||
* | From 0 to 1 spare | Jon Bratseth | 2020-06-23 | 13 | -19/+57 |
| | |||||
* | Enable configserver-provision-lb by default | Martin Polden | 2020-06-22 | 1 | -2/+2 |
| | |||||
* | Provision load balancer for controller cluster | Martin Polden | 2020-06-19 | 1 | -0/+15 |
| | |||||
* | Merge pull request #13619 from vespa-engine/bratseth/spare-capacity-maintainer | Jon Bratseth | 2020-06-18 | 15 | -38/+368 |
|\ | | | | | Bratseth/spare capacity maintainer | ||||
| * | Less max iterations as each do more exporation | Jon Bratseth | 2020-06-17 | 1 | -2/+2 |
| | | |||||
| * | Limit by iterations instead of depth | Jon Bratseth | 2020-06-17 | 1 | -3/+7 |
| | | |||||
| * | Memoize | Jon Bratseth | 2020-06-17 | 1 | -0/+25 |
| | | |||||
| * | More tests | Jon Bratseth | 2020-06-17 | 1 | -6/+20 |
| | | |||||
| * | More tests | Jon Bratseth | 2020-06-17 | 1 | -1/+71 |
| | | |||||
| * | Test SpareCapacityMaintainer | Jon Bratseth | 2020-06-17 | 3 | -9/+226 |
| | | |||||
| * | SpareCapacityMaintainer sketch | Jon Bratseth | 2020-06-12 | 12 | -26/+26 |
| | | |||||
| * | Spare capacity maintainenance skeleton | Jon Bratseth | 2020-06-09 | 1 | -3/+3 |
| | | |||||
* | | Provision load balancer for config server cluster | Martin Polden | 2020-06-11 | 1 | -1/+19 |
| | | |||||
* | | Extract LoadBalancerSpec | Martin Polden | 2020-06-11 | 2 | -4/+4 |
| | | |||||
* | | Remove NodeRepository from LoadBalancerService interface | Martin Polden | 2020-06-11 | 2 | -4/+4 |
| | | |||||
* | | Revert "Provision load balancer for config server cluster" | Martin Polden | 2020-06-11 | 3 | -24/+6 |
| | | |||||
* | | Provision load balancer for config server cluster | Martin Polden | 2020-06-11 | 1 | -1/+19 |
| | | |||||
* | | Extract LoadBalancerSpec | Martin Polden | 2020-06-11 | 2 | -4/+4 |
| | |