Merge pull request #22460 from vespa-engine/olaa/cmr-ignore-failed-hosts

Failed and breakfixed hosts are always noop in CMRs

1 files changed, 9 insertions, 0 deletions

diff --git a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/maintenance/VcmrMaintainer.java b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/maintenance/VcmrMaintainer.java
index 1463cce595d..551f803f368 100644
--- a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/maintenance/VcmrMaintainer.java
+++ b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/maintenance/VcmrMaintainer.java
@@ -214,6 +214,10 @@ public class VcmrMaintainer extends ControllerMaintainer {
             return hostAction.withState(State.PENDING_RETIREMENT);
         }
 
+        if (isFailed(node)) {
+            return hostAction.withState(State.NONE);
+        }
+
         return hostAction;
     }
 
@@ -260,6 +264,11 @@ public class VcmrMaintainer extends ControllerMaintainer {
                 action.getState() == State.RETIRING && !node.wantToRetire();
     }
 
+    private boolean isFailed(Node node) {
+        return node.state() == Node.State.failed ||
+                node.state() == Node.State.breakfixed;
+    }
+
     private Map<ZoneId, List<Node>> nodesByZone() {
         return controller().zoneRegistry()
                 .zones()