From 1e5d534c3ca32c9c775613dddae12fcfab15e2bf Mon Sep 17 00:00:00 2001
From: dahn <daan@onecht.net>
Date: Mon, 17 Aug 2026 09:42:54 +0200
Subject: [PATCH] guard against stopping stats background threads (#13659)

* guard against stopping stats background threats

* stats cleanup test

* override annotation, and specic and generic enough catch

---------

Co-authored-by: Daan Hoogland <dahn@apache.org>
---
 .../java/com/cloud/server/StatsCollector.java | 14 +++++-
 .../com/cloud/server/StatsCollectorTest.java  | 43 +++++++++++++++++++
 2 files changed, 55 insertions(+), 2 deletions(-)

diff --git a/server/src/main/java/com/cloud/server/StatsCollector.java b/server/src/main/java/com/cloud/server/StatsCollector.java
index 1e0138f7cf9..ca8d95c74c5 100644
--- a/server/src/main/java/com/cloud/server/StatsCollector.java
+++ b/server/src/main/java/com/cloud/server/StatsCollector.java
@@ -1285,14 +1285,24 @@ public class StatsCollector extends ManagerBase implements ComponentMethodInterc
      * can be enabled/disabled independently.</p>
      */
     class VmStatsCleaner extends ManagedContextRunnable{
+        @Override
         protected void runInContext() {
-            cleanUpVirtualMachineStats();
+            try {
+                cleanUpVirtualMachineStats();
+            } catch (RuntimeException e) {
+                logger.error("Error trying to clean up VM stats", e);
+            }
         }
     }
 
     class VolumeStatsCleaner extends ManagedContextRunnable{
+        @Override
         protected void runInContext() {
-            cleanUpVolumeStats();
+            try {
+                cleanUpVolumeStats();
+            } catch (RuntimeException e) {
+                logger.error("Error trying to clean up Volume stats", e);
+            }
         }
     }
 
diff --git a/server/src/test/java/com/cloud/server/StatsCollectorTest.java b/server/src/test/java/com/cloud/server/StatsCollectorTest.java
index 3578e6948a4..46449a9cb6e 100644
--- a/server/src/test/java/com/cloud/server/StatsCollectorTest.java
+++ b/server/src/test/java/com/cloud/server/StatsCollectorTest.java
@@ -335,6 +335,49 @@ public class StatsCollectorTest {
         Mockito.verify(vmStatsDaoMock).removeAllByTimestampLessThan(Mockito.any(), Mockito.anyLong());
     }
 
+    private void setVmDiskStatsMaxRetentionTimeValue(String value) {
+        StatsCollector.vmDiskStatsMaxRetentionTime = new ConfigKey<Integer>("Advanced", Integer.class, "vm.disk.stats.max.retention.time", value,
+                "The maximum time (in minutes) for keeping Volume stats records in the database. The Volume stats cleanup process will be disabled if this is set to 0 or less than 0.", true);
+    }
+
+    @Test
+    public void cleanUpVolumeStatsTestIsDisabled() {
+        setVmDiskStatsMaxRetentionTimeValue("0");
+
+        statsCollector.cleanUpVolumeStats();
+
+        Mockito.verify(volumeStatsDao, Mockito.never()).removeAllByTimestampLessThan(Mockito.any(), Mockito.anyLong());
+    }
+
+    @Test
+    public void cleanUpVolumeStatsTestIsEnabled() {
+        setVmDiskStatsMaxRetentionTimeValue("1");
+
+        statsCollector.cleanUpVolumeStats();
+
+        Mockito.verify(volumeStatsDao).removeAllByTimestampLessThan(Mockito.any(), Mockito.anyLong());
+    }
+
+    @Test
+    public void vmStatsCleanerTestCatchesCloudRuntimeExceptionAndKeepsRunning() {
+        Mockito.doThrow(new CloudRuntimeException("Communications link failure")).when(statsCollector).cleanUpVirtualMachineStats();
+        StatsCollector.VmStatsCleaner vmStatsCleaner = statsCollector.new VmStatsCleaner();
+
+        vmStatsCleaner.run();
+
+        Mockito.verify(statsCollector).cleanUpVirtualMachineStats();
+    }
+
+    @Test
+    public void volumeStatsCleanerTestCatchesCloudRuntimeExceptionAndKeepsRunning() {
+        Mockito.doThrow(new CloudRuntimeException("Communications link failure")).when(statsCollector).cleanUpVolumeStats();
+        StatsCollector.VolumeStatsCleaner volumeStatsCleaner = statsCollector.new VolumeStatsCleaner();
+
+        volumeStatsCleaner.run();
+
+        Mockito.verify(statsCollector).cleanUpVolumeStats();
+    }
+
     @Test
     public void persistVirtualMachineStatsTestPersistsSuccessfully() {
         statsCollector.msId = 1L;