Merge pull request #12869 from vespa-engine/jvenstad/all-jobs-in-deployment-api

Provide all (relevant runs of all) jobs in deployment API

3 files changed, 308 insertions, 28 deletions

diff --git a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiHandler.java b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiHandler.java
index 4f2092a590c..ec1ff1ebc99 100644
--- a/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiHandler.java
+++ b/controller-server/src/main/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiHandler.java
@@ -1,7 +1,6 @@
 // Copyright 2019 Oath Inc. Licensed under the terms of the Apache 2.0 license. See LICENSE in the project root.
 package com.yahoo.vespa.hosted.controller.restapi.deployment;
 
-import com.yahoo.component.Version;
 import com.yahoo.component.Vtag;
 import com.yahoo.config.application.api.DeploymentInstanceSpec;
 import com.yahoo.config.application.api.DeploymentSpec;
@@ -17,29 +16,25 @@ import com.yahoo.restapi.Uri;
 import com.yahoo.slime.Cursor;
 import com.yahoo.slime.Slime;
 import com.yahoo.vespa.hosted.controller.Controller;
+import com.yahoo.vespa.hosted.controller.api.integration.deployment.JobType;
 import com.yahoo.vespa.hosted.controller.application.ApplicationList;
 import com.yahoo.vespa.hosted.controller.application.TenantAndApplicationId;
-import com.yahoo.vespa.hosted.controller.deployment.DeploymentStatus;
-import com.yahoo.vespa.hosted.controller.deployment.DeploymentStatusList;
-import com.yahoo.vespa.hosted.controller.deployment.JobList;
 import com.yahoo.vespa.hosted.controller.deployment.Run;
-import com.yahoo.vespa.hosted.controller.deployment.RunStatus;
 import com.yahoo.vespa.hosted.controller.restapi.application.EmptyResponse;
 import com.yahoo.vespa.hosted.controller.versions.DeploymentStatistics;
 import com.yahoo.vespa.hosted.controller.versions.VespaVersion;
 import com.yahoo.yolean.Exceptions;
 
-import java.time.Instant;
-import java.util.Comparator;
-import java.util.Map;
-import java.util.Optional;
-import java.util.Set;
-import java.util.function.Function;
+import java.util.LinkedHashMap;
+import java.util.List;
 import java.util.logging.Level;
 import java.util.stream.Collectors;
+import java.util.stream.Stream;
 
+import static java.util.function.Function.identity;
 import static java.util.stream.Collectors.groupingBy;
-import static java.util.stream.Collectors.toUnmodifiableList;
+import static java.util.stream.Collectors.toList;
+import static java.util.stream.Collectors.toMap;
 import static java.util.stream.Collectors.toUnmodifiableMap;
 
 /**
@@ -99,9 +94,9 @@ public class DeploymentApiHandler extends LoggingRequestHandler {
         var versionStatus = controller.versionStatus();
         var systemVersion = versionStatus.systemVersion().map(VespaVersion::versionNumber).orElse(Vtag.currentVersion);
         var deploymentStatuses = controller.jobController().deploymentStatuses(ApplicationList.from(controller.applications().asList()), systemVersion);
-        var deploymentStatistics = DeploymentStatistics.compute(versionStatus.versions().stream().map(VespaVersion::versionNumber).collect(Collectors.toList()),
+        var deploymentStatistics = DeploymentStatistics.compute(versionStatus.versions().stream().map(VespaVersion::versionNumber).collect(toList()),
                                                                 deploymentStatuses)
-                                                       .stream().collect(Collectors.toMap(DeploymentStatistics::version, Function.identity()));
+                                                       .stream().collect(toMap(DeploymentStatistics::version, identity()));
         for (VespaVersion version : versionStatus.versions()) {
             Cursor versionObject = platformArray.addObject();
             versionObject.setString("version", version.versionNumber().toString());
@@ -117,8 +112,9 @@ public class DeploymentApiHandler extends LoggingRequestHandler {
                 configServerObject.setString("hostname", hostname.value());
             }
 
+            DeploymentStatistics statistics = deploymentStatistics.get(version.versionNumber());
             Cursor failingArray = versionObject.setArray("failingApplications");
-            for (Run run : deploymentStatistics.get(version.versionNumber()).failingUpgrades()) {
+            for (Run run : statistics.failingUpgrades()) {
                 Cursor applicationObject = failingArray.addObject();
                 toSlime(applicationObject, run.id().application(), request);
                 applicationObject.setString("failing", run.id().type().jobName());
@@ -129,25 +125,76 @@ public class DeploymentApiHandler extends LoggingRequestHandler {
                                                    .flatMap(status -> status.instanceJobs().entrySet().stream())
                                                    .collect(toUnmodifiableMap(jobs -> jobs.getKey(), jobs -> jobs.getValue()));
             Cursor productionArray = versionObject.setArray("productionApplications");
-            deploymentStatistics.get(version.versionNumber()).productionSuccesses().stream()
-                                .collect(groupingBy(run -> run.id().application()))
-                                .forEach((id, runs) -> {
-                                    Cursor applicationObject = productionArray.addObject();
-                                    toSlime(applicationObject, id, request);
-                                    applicationObject.setLong("productionJobs", jobsByInstance.get(id).production().size());
-                                    applicationObject.setLong("productionSuccesses", runs.size());
-                                });
+            statistics.productionSuccesses().stream()
+                      .collect(groupingBy(run -> run.id().application()))
+                      .forEach((id, runs) -> {
+                          Cursor applicationObject = productionArray.addObject();
+                          toSlime(applicationObject, id, request);
+                          applicationObject.setLong("productionJobs", jobsByInstance.get(id).production().size());
+                          applicationObject.setLong("productionSuccesses", runs.size());
+                      });
 
             Cursor runningArray = versionObject.setArray("deployingApplications");
-            for (Run run : deploymentStatistics.get(version.versionNumber()).runningUpgrade()) {
+            for (Run run : statistics.runningUpgrade()) {
                 Cursor applicationObject = runningArray.addObject();
                 toSlime(applicationObject, run.id().application(), request);
                 applicationObject.setString("running", run.id().type().jobName());
             }
+
+            class RunInfo { //  ヽ༼ຈل͜ຈ༽━☆ﾟ.*･｡ﾟ
+                final Run run;
+                final boolean upgrade;
+                RunInfo(Run run, boolean upgrade) { this.run = run; this.upgrade = upgrade; }
+                @Override public String toString() { return run.id().toString(); }
+            }
+            Cursor instancesArray = versionObject.setArray("applications");
+            Stream.of(statistics.failingUpgrades().stream().map(run -> new RunInfo(run, true)),
+                      statistics.otherFailing().stream().map(run -> new RunInfo(run, false)),
+                      statistics.runningUpgrade().stream().map(run -> new RunInfo(run, true)),
+                      statistics.otherRunning().stream().map(run -> new RunInfo(run, false)),
+                      statistics.productionSuccesses().stream().map(run -> new RunInfo(run, true)))
+                  .flatMap(identity())
+                  .collect(Collectors.groupingBy(run -> run.run.id().application(),
+                                                 LinkedHashMap::new, // Put apps with failing and running jobs first.
+                                                 groupingBy(run -> run.run.id().type(),
+                                                            LinkedHashMap::new,
+                                                            toList())))
+                  .forEach((instance, runs) -> {
+                      Cursor instanceObject = instancesArray.addObject();
+                      instanceObject.setString("tenant", instance.tenant().value());
+                      instanceObject.setString("application", instance.application().value());
+                      instanceObject.setString("instance", instance.instance().value());
+                      instanceObject.setLong("productionJobCount", jobsByInstance.get(instance).production().size());
+                      instanceObject.setString("upgradePolicy", toString(deploymentStatuses.matching(status -> status.application().id().equals(TenantAndApplicationId.from(instance)))
+                                                                                           .first().map(status -> status.application().deploymentSpec())
+                                                                                           .flatMap(spec -> spec.instance(instance.instance()).map(DeploymentInstanceSpec::upgradePolicy))
+                                                                                           .orElse(DeploymentSpec.UpgradePolicy.defaultPolicy)));
+                      Cursor allJobsObject = instanceObject.setObject("allJobs");
+                      Cursor upgradeJobsObject = instanceObject.setObject("upgradeJobs");
+                      runs.forEach((type, rs) -> {
+                          Cursor jobObject = allJobsObject.setObject(type.jobName());
+                          Cursor upgradeObject = upgradeJobsObject.setObject(type.jobName());
+                          for (RunInfo run : rs) {
+                              toSlime(jobObject, run.run);
+                              if (run.upgrade)
+                                  toSlime(upgradeObject, run.run);
+                          }
+                      });
+                  });
         }
+        JobType.allIn(controller.system()).stream().map(JobType::jobName).forEach(root.setArray("jobs")::addString);
         return new SlimeJsonResponse(slime);
     }
 
+    private void toSlime(Cursor jobObject, Run run) {
+        String key = run.hasFailed() ? "failing" : run.hasEnded() ? "success" : "running";
+        Cursor runObject = jobObject.setObject(key);
+        runObject.setLong("number", run.id().number());
+        runObject.setLong("start", run.start().toEpochMilli());
+        run.end().ifPresent(end -> runObject.setLong("end", end.toEpochMilli()));
+        runObject.setString("status", run.status().name());
+    }
+
     private void toSlime(Cursor object, ApplicationId id, HttpRequest request) {
         object.setString("tenant", id.tenant().value());
         object.setString("application", id.application().value());
@@ -168,6 +215,4 @@ public class DeploymentApiHandler extends LoggingRequestHandler {
         return upgradePolicy.name();
     }
 
-    // ----------------------------- Utilities to pick out the relevant JobStatus -- filter chains should mirror the ones in VersionStatus
-
 }
diff --git a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiTest.java b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiTest.java
index 2118f221a79..c8036676fa9 100644
--- a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiTest.java
+++ b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/DeploymentApiTest.java
@@ -69,6 +69,9 @@ public class DeploymentApiTest extends ControllerContainerTest {
         deploymentTester.upgrader().maintain();
         deploymentTester.triggerJobs();
 
+        // Application fails application change
+        productionApp.submit(multiInstancePackage).failDeployment(JobType.systemTest);
+
         tester.controller().updateVersionStatus(censorConfigServers(VersionStatus.compute(tester.controller())));
         tester.assertResponse(authenticatedRequest("http://localhost:8080/deployment/v1/"), new File("root.json"));
         tester.assertResponse(authenticatedRequest("http://localhost:8080/api/deployment/v1/"), new File("root.json"));
diff --git a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/responses/root.json b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/responses/root.json
index 50144b725e6..f44c3bce8a8 100644
--- a/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/responses/root.json
+++ b/controller-server/src/test/java/com/yahoo/vespa/hosted/controller/restapi/deployment/responses/root.json
@@ -29,7 +29,63 @@
           "productionSuccesses": 1
         }
       ],
-      "deployingApplications": []
+      "deployingApplications": [],
+      "applications": [
+        {
+          "tenant": "tenant1",
+          "application": "application1",
+          "instance": "default",
+          "productionJobCount": 1,
+          "upgradePolicy": "default",
+          "allJobs": {
+            "production-us-west-1": {
+              "success": {
+                "number": 1,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          },
+          "upgradeJobs": {
+            "production-us-west-1": {
+              "success": {
+                "number": 1,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          }
+        },
+        {
+          "tenant": "tenant2",
+          "application": "application2",
+          "instance": "i2",
+          "productionJobCount": 1,
+          "upgradePolicy": "default",
+          "allJobs": {
+            "production-us-west-1": {
+              "success": {
+                "number": 1,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          },
+          "upgradeJobs": {
+            "production-us-west-1": {
+              "success": {
+                "number": 1,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          }
+        }
+      ]
     },
     {
       "version": "5.1",
@@ -102,8 +158,184 @@
           "upgradePolicy": "default",
           "running": "production-us-west-1"
         }
+      ],
+      "applications": [
+        {
+          "tenant": "tenant1",
+          "application": "application1",
+          "instance": "default",
+          "productionJobCount": 1,
+          "upgradePolicy": "default",
+          "allJobs": {
+            "system-test": {
+              "failing": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "error"
+              },
+              "running": {
+                "number": 3,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "staging-test": {
+              "failing": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "error"
+              },
+              "running": {
+                "number": 3,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            }
+          },
+          "upgradeJobs": {
+            "system-test": {
+              "failing": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "error"
+              },
+              "running": {
+                "number": 3,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "staging-test": {
+              "failing": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "error"
+              },
+              "running": {
+                "number": 3,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            }
+          }
+        },
+        {
+          "tenant": "tenant2",
+          "application": "application2",
+          "instance": "i1",
+          "productionJobCount": 1,
+          "upgradePolicy": "default",
+          "allJobs": {
+            "system-test": {
+              "failing": {
+                "number": 3,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "error"
+              }
+            },
+            "staging-test": {
+              "running": {
+                "number": 3,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "production-us-west-1": {
+              "success": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          },
+          "upgradeJobs": {
+            "system-test": {},
+            "staging-test": {},
+            "production-us-west-1": {
+              "success": {
+                "number": 2,
+                "start": "(ignore)",
+                "end": "(ignore)",
+                "status": "success"
+              }
+            }
+          }
+        },
+        {
+          "tenant": "tenant2",
+          "application": "application2",
+          "instance": "i2",
+          "productionJobCount": 1,
+          "upgradePolicy": "default",
+          "allJobs": {
+            "production-us-west-1": {
+              "running": {
+                "number": 2,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "system-test": {
+              "running": {
+                "number": 1,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "staging-test": {
+              "running": {
+                "number": 1,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            }
+          },
+          "upgradeJobs": {
+            "production-us-west-1": {
+              "running": {
+                "number": 2,
+                "start": "(ignore)",
+                "status": "running"
+              }
+            },
+            "system-test": {},
+            "staging-test": {}
+          }
+        }
       ]
     }
+  ],
+  "jobs": [
+    "system-test",
+    "staging-test",
+    "production-us-east-3",
+    "test-us-east-3",
+    "production-us-west-1",
+    "test-us-west-1",
+    "production-us-central-1",
+    "test-us-central-1",
+    "production-ap-northeast-1",
+    "test-ap-northeast-1",
+    "production-ap-northeast-2",
+    "test-ap-northeast-2",
+    "production-ap-southeast-1",
+    "test-ap-southeast-1",
+    "production-eu-west-1",
+    "test-eu-west-1",
+    "production-aws-us-east-1a",
+    "test-aws-us-east-1a",
+    "production-aws-us-west-2a",
+    "test-aws-us-west-2a",
+    "production-aws-us-east-1b",
+    "test-aws-us-east-1b",
+    "dev-us-east-1",
+    "dev-aws-us-east-2a",
+    "perf-us-east-3"
   ]
 }
-