Merge pull request #26712 from vespa-engine/jonmv/avoid-false-broken-confidence

When a canary fails after other apps upgrade that is not broken confi…

2 files changed, 9 insertions, 2 deletions

diff --git a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/versions/VespaVersion.java b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/versions/VespaVersion.java
index e9947f3d565..45c00848407 100644
--- a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/versions/VespaVersion.java
+++ b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/versions/VespaVersion.java
@@ -41,8 +41,8 @@ public record VespaVersion(Version version,
                                            .not().upgradingTo(statistics.version());
         InstanceList failingOnThis = all.matching(instance -> statistics.failingUpgrades().stream().anyMatch(run -> run.id().application().equals(instance)));
 
-        // 'broken' if any canary fails
-        if  ( ! failingOnThis.with(UpgradePolicy.canary).isEmpty())
+        // 'broken' if any canary fails, and no non-canary is upgraded
+        if  ( ! failingOnThis.with(UpgradePolicy.canary).isEmpty() && productionOnThis.not().with(UpgradePolicy.canary).isEmpty())
             return Confidence.broken;
 
         // 'broken' if 6 non-canary was broken by this, and that is at least 5% of all
diff --git a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/versions/VersionStatusTest.java b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/versions/VersionStatusTest.java
index f08e92a515d..7afa5c7f44a 100644
--- a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/versions/VersionStatusTest.java
+++ b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/versions/VersionStatusTest.java
@@ -349,6 +349,13 @@ public class VersionStatusTest {
         assertEquals(Confidence.high, confidence(tester.controller(), version0), "Confidence remains unchanged for version0: High");
         assertEquals(VespaVersion.Confidence.high, confidence(tester.controller(), version2), "90% of defaults deployed successfully: High");
 
+        // Canary failing a new revision does not affect confidence
+        canary0.submit(canaryPolicy).failDeployment(systemTest);
+        tester.controllerTester().computeVersionStatus();
+        assertEquals(Confidence.high, confidence(tester.controller(), version0), "Confidence remains unchanged for version0: High");
+        assertEquals(VespaVersion.Confidence.high, confidence(tester.controller(), version2), "90% of defaults deployed successfully: High");
+        canary0.deploy();
+
         // A new version is released, all canaries upgrade successfully, but enough "default" apps fail to mark version
         // as broken
         Version version3 = new Version("6.5");