Merge "Emit stats for more build results"

releasenotes/notes/build-result-stats-5e80a82a78faeb61.yaml

@@ -0,0 +1,7 @@
+---
+features:
+  - |
+    Monitoring stats for builds are now emitted for more results.
+    Previously they were emitted for some build results such as
+    `SUCCESS`, `FAILURE`, and `ERROR` but now include `NODE_FAILURE`
+    and other result values.

tests/unit/test_scheduler.py

@@ -5933,6 +5933,18 @@ For CI problems and help debugging, contact ci@example.org"""
                                   A.messages[1]))
         self.assertTrue(
             'Skipped due to failed job project-merge' in A.messages[1])
+        self.assertReportedStat(
+            'zuul.nodepool.requests.requested.total', value='1', kind='c')
+        self.assertReportedStat(
+            'zuul.nodepool.requests.requested.label.label1',
+            value='1', kind='c')
+        self.assertReportedStat(
+            'zuul.nodepool.requests.failed.label.label1',
+            value='1', kind='c')
+        self.assertReportedStat(
+            'zuul.tenant.tenant-one.pipeline.gate.project.review_example_com.'
+            'org_project.master.job.project-merge.NODE_FAILURE', value='1',
+            kind='c')
 
     def test_nodepool_resources(self):
         "Test that resources are reported"

zuul/model.py

@@ -5472,7 +5472,7 @@ class QueueItem(zkobject.ZKObject):
                                   job=job, build_set=self.current_build_set,
                                   error_detail=str(e), result='FAILURE')
             self.addBuild(fakebuild)
-            self.pipeline.manager.sql.reportBuildEnd(
+            self.pipeline.manager.sched.reportBuildEnd(
                 fakebuild,
                 tenant=self.pipeline.tenant.name,
                 final=True)
@@ -5881,7 +5881,7 @@ class QueueItem(zkobject.ZKObject):
                                       error_detail=skipped_reason,
                                       result='SKIPPED')
                 self.addBuild(fakebuild)
-                self.pipeline.manager.sql.reportBuildEnd(
+                self.pipeline.manager.sched.reportBuildEnd(
                     fakebuild,
                     tenant=self.pipeline.tenant.name,
                     final=True)
@@ -5897,7 +5897,7 @@ class QueueItem(zkobject.ZKObject):
             result='NODE_FAILURE',
         )
         self.addBuild(fakebuild)
-        self.pipeline.manager.sql.reportBuildEnd(
+        self.pipeline.manager.sched.reportBuildEnd(
             fakebuild,
             tenant=self.pipeline.tenant.name,
             final=True)
@@ -5958,7 +5958,7 @@ class QueueItem(zkobject.ZKObject):
                                   job=job, build_set=self.current_build_set,
                                   error_detail=msg, result='SKIPPED')
             self.addBuild(fakebuild)
-            self.pipeline.manager.sql.reportBuildEnd(
+            self.pipeline.manager.sched.reportBuildEnd(
                 fakebuild,
                 tenant=self.pipeline.tenant.name,
                 final=True)
@@ -5972,7 +5972,7 @@ class QueueItem(zkobject.ZKObject):
                                   job=job, build_set=self.current_build_set,
                                   error_detail=msg, result='SKIPPED')
             self.addBuild(fakebuild)
-            self.pipeline.manager.sql.reportBuildEnd(
+            self.pipeline.manager.sched.reportBuildEnd(
                 fakebuild,
                 tenant=self.pipeline.tenant.name,
                 final=True)

zuul/scheduler.py

@@ -887,6 +887,10 @@ class Scheduler(threading.Thread):
             event, needs_result=False
         )
 
+    def reportBuildEnd(self, build, tenant, final):
+        self.sql.reportBuildEnd(build, tenant=tenant, final=final)
+        self._reportBuildStats(build)
+
     def _reportBuildStats(self, build):
         # Note, as soon as the result is set, other threads may act
         # upon this, even though the event hasn't been fully
@@ -2904,6 +2908,9 @@ class Scheduler(threading.Thread):
             )
 
             self._cleanupCompletedBuild(build)
+            # This usually happens when we have already canceled a
+            # build and reported stats for it, so don't send stats in
+            # this case.
             try:
                 self.sql.reportBuildEnd(
                     build, tenant=pipeline.tenant.name,
@@ -2974,11 +2981,9 @@ class Scheduler(threading.Thread):
             }
             tracing.endSavedSpan(build.span_info, attributes=attributes)
 
-        self._reportBuildStats(build)
-
         self._cleanupCompletedBuild(build)
         try:
-            self.sql.reportBuildEnd(
+            self.reportBuildEnd(
                 build, tenant=pipeline.tenant.name, final=(not build.retry))
         except Exception:
             log.exception("Error reporting build completion to DB:")
@@ -3130,7 +3135,7 @@ class Scheduler(threading.Thread):
                     # anymore once the pipeline is removed.
                     self.executor.removeBuild(build)
                     try:
-                        self.sql.reportBuildEnd(
+                        self.reportBuildEnd(
                             build, build.build_set.item.pipeline.tenant.name,
                             final=False)
                     except Exception: