aboutsummaryrefslogtreecommitdiffstats
path: root/node-repository/src/main/java
Commit message (Collapse)AuthorAgeFilesLines
* Find current load more reliablyJon Bratseth2022-07-142-4/+17
|
* Make smaller resource changesJon Bratseth2022-07-141-1/+1
|
* No functional changesJon Bratseth2022-07-142-2/+3
|
* Always include deferOsUpgrade field for hostsMartin Polden2022-07-131-3/+3
|
* Merge pull request #23474 from ↵Håkon Hallingstad2022-07-121-1/+1
|\ | | | | | | | | vespa-engine/hakonhall/run-nodefailer-every-3m-instead-of-5m Run NodeFailer every 3m instead of 5m
| * Run NodeFailer every 3m instead of 5mHåkon Hallingstad2022-07-121-1/+1
| |
* | Add total cost to statsJon Bratseth2022-07-122-3/+16
| |
* | Allow failed node to be deprovisionedMartin Polden2022-07-121-1/+2
| |
* | Node with allocation must be parked to allow deprovisioningMartin Polden2022-07-121-16/+29
| |
* | Reapply "Allow deprovision of parked host & node w/alloc when node has ↵Martin Polden2022-07-121-1/+1
|/ | | | | | wantToDeprovision" This reverts commit d097cb3bf2808bb05f2dc4fc2e7cf771246ba1a9.
* Only use wantToFail just before activateHåkon Hallingstad2022-07-111-5/+10
|
* Revert "Revert "Avoid the host lock while failing the children""Håkon Hallingstad2022-07-111-46/+54
|
* Revert "Avoid the host lock while failing the children"Håkon Hallingstad2022-07-111-54/+46
| | | | This reverts commit 2cdaef56e18ace2ee2269d28f959f5a534bd68ee.
* Revert update of commentHåkon Hallingstad2022-07-111-2/+2
|
* Merge pull request #23440 from ↵v8.15.63Harald Musum2022-07-081-2/+2
|\ | | | | | | | | vespa-engine/hakonhall/define-main-chain-graph-flag Define main-chain-graph flag
| * Define main-chain-graph flagHåkon Hallingstad2022-07-081-2/+2
| |
* | Reduce scope of unallocated lock and avoid deadlockMartin Polden2022-07-081-21/+15
| | | | | | | | | | | | Before this change a call to `failOrMarkRecursively` could cause a deadlock because we would then take the application lock while holding unallocatedLock, but a deployment (e.g. by `InfrastructureProvisioner`) does the opposite.
* | Add deferOsUpgrade field to node responseMartin Polden2022-07-082-6/+15
| |
* | Limit grace period to RetiringOsUpgraderMartin Polden2022-07-084-14/+20
| |
* | Add a grace period before upgrading new nodesJon Bratseth2022-07-076-0/+23
|/
* Avoid the host lock while failing the childrenHåkon Hallingstad2022-07-051-46/+54
|
* Autoscaling should happen within 5 minutesJon Bratseth2022-07-051-1/+1
|
* Merge pull request #23345 from ↵Harald Musum2022-07-041-1/+1
|\ | | | | | | | | vespa-engine/hmusum/use-correct-agent-when-failing-nodes Use correct agent when failing nodes
| * Use correct agent when failing nodesHarald Musum2022-07-041-1/+1
| |
* | Update javadoc and reduce log levelHarald Musum2022-07-041-3/+4
|/
* Reuse fully retired nodes fasterMartin Polden2022-06-289-60/+98
|
* Merge pull request #23164 from vespa-engine/hmusum/add-getActivatedTimeJon Bratseth2022-06-201-1/+1
|\ | | | | Use getActivatedTime() for last deployed time for an app [run-systemtest]
| * Add getActivatedTime() for a sessionHarald Musum2022-06-201-1/+1
| | | | | | | | | | | | Use getActivatedTime() instead of getCreatedTime in lastDeployTime(). getCreatedTime() gives time a new session was created, not when it was activated, which is what we usually want.
* | Revert "Allow deprovision of parked host & node w/alloc when node has ↵Martin Polden2022-06-201-1/+1
|/ | | | wantToDeprovision"
* Reduce interval for DynamicProvisioningMaintainer to 3 minutesHarald Musum2022-06-181-1/+1
| | | | | Waiting for provisinong hosts takes a really long time, resuming provisioning more often might help a little
* Merge pull request #23107 from vespa-engine/freva/do-not-clear-wtdHåkon Hallingstad2022-06-151-0/+1
|\ | | | | Do not reset node status if wantToDeprovision
| * Do not reset node status if wantToDeprovisionValerij Fredriksen2022-06-151-0/+1
| |
* | Remove cluster from autoscaling advice messagesHarald Musum2022-06-151-4/+4
|/
* Cosmetix fix, id.toString() already contains "cluster "Harald Musum2022-06-142-5/+3
|
* Merge pull request #23061 from ↵Henning Baldersheim2022-06-131-2/+6
|\ | | | | | | | | vespa-engine/hmusum/add-application-id-to-illegal-argument-exception Add application id to IllegalArgumentException in AutoscalingMaintainer
| * Chain exceptionsHarald Musum2022-06-131-3/+1
| |
| * Add application id to IllegalArgumentException in AutoscalingMaintainerHarald Musum2022-06-131-1/+7
| |
* | More info in autoscaling adviceHarald Musum2022-06-131-5/+7
|/ | | | | Hard to debug why autoscaling does or does not happen, add some more info
* Deprovision host with host lock and sanity-checkHåkon Hallingstad2022-06-091-6/+12
|
* Remove '.sum' form vds sum metrics.Henning Baldersheim2022-06-081-3/+3
| | | | | | | | Remove '.sum' from metric names for storage node and also remove the average metrics for the same. Remove '.sum' from distributor metrics set and remove distributor average metrics. GC '.sum' from distributor metric names. Remove '.alldisks' from metric names and update tests. GC '.alldisks' from filestor metrics.
* Mark host as wantToDeprovision before deprovisioningValerij Fredriksen2022-06-071-1/+5
|
* Remove cloud account restrictionMartin Polden2022-06-021-4/+0
|
* Implement HostRetirerMartin Polden2022-06-014-1/+86
|
* Allow deprovision of parked host & node w/alloc when node has wantToDeprovisionHåkon Hallingstad2022-06-011-1/+1
|
* Define smallest node resources that will work in GCPHarald Musum2022-05-281-14/+29
|
* Add some debug logging when provisoning nodes for a clusterHarald Musum2022-05-281-0/+5
| | | | Helps finding cause of provisoning failures
* Allow patching wantedOsVersionMartin Polden2022-05-203-4/+9
|
* Choose node resources with a matching host flavor when exclusiveMartin Polden2022-05-193-20/+33
| | | | | | | | When using a custom cloud account (always exclusive) we cannot choose a too small flavor because there may not be any matching host flavor. This currently works in our own zones because there is always a shared host that can be used for admin nodes (feature flag is set in all zones) and there is no way to set exclusivity requirement for those clusters.
* Never downsize if allocating exclusivelyMartin Polden2022-05-182-9/+9
|
* Let CapacityPolicies decide exclusivity based on cloud accountMartin Polden2022-05-183-2/+6
|