roachtest: replace kv/splits with splits perturbation test (#170090)

roachtest: replace kv/splits with splits perturbation test

Author: trunk-io[bot]

pkg/cmd/roachtest/tests/kv.go

@@ -767,105 +767,6 @@ func drainWithIpTables(
 	c.Start(ctx, t.L(), startOpts, install.MakeClusterSettings(), restartNode)
 }
 
-func registerKVSplits(r registry.Registry) {
-	for _, item := range []struct {
-		quiesce bool
-		splits  int
-		leases  registry.LeaseType
-		timeout time.Duration
-		envVars []string
-	}{
-		// NB: with 500000 splits, this test sometimes fails since it's pushing
-		// far past the number of replicas per node we support, at least if the
-		// ranges start to unquiesce (which can set off a cascade due to resource
-		// exhaustion).
-		{true, 200_000, registry.EpochLeases, 140 * time.Minute, nil},
-		// This version of the test prevents range quiescence to trigger the
-		// badness described above more reliably for when we wish to improve
-		// the performance. For now, just verify that 30k unquiesced ranges
-		// is tenable.
-		{false, 30_000, registry.EpochLeases, 2 * time.Hour, nil},
-		// Expiration-based leases prevent quiescence, and are also more expensive
-		// to keep alive. Again, just verify that 30k ranges is ok.
-		{false, 30_000, registry.ExpirationLeases, 2 * time.Hour, nil},
-		// Leader leases without quiescence perform similarly to epoch leases
-		// without quiescence. Even though epoch leases are set to reach 30k, they
-		// also reach 60k reliably.
-		{false, 60_000, registry.LeaderLeases, 2 * time.Hour, nil},
-		// Leader leases with quiescence don't quite match epoch leases with
-		// quiescence because in leader leases only the followers ever quiesce.
-		{true, 80_000, registry.LeaderLeases, 2 * time.Hour, nil},
-		// With some additional tuning, leader leases can do even better. The extended interval allow
-		// for more flexibility in extending store liveness support, and prevent support withdrawals at
-		// higher CPU utilization when goroutine scheduling latency is high.
-		{
-			true, 100_000, registry.LeaderLeases, 2 * time.Hour,
-			[]string{
-				"COCKROACH_STORE_LIVENESS_SUPPORT_EXPIRY_INTERVAL=1s",
-				"COCKROACH_STORE_LIVENESS_HEARTBEAT_INTERVAL=3s",
-				"COCKROACH_STORE_LIVENESS_SUPPORT_DURATION=6s",
-			},
-		},
-	} {
-		item := item // for use in closure below
-		name := fmt.Sprintf("kv/splits/nodes=3/quiesce=%t/lease=%s", item.quiesce, item.leases)
-		if item.envVars != nil {
-			name += "/tuned"
-		}
-		r.Add(registry.TestSpec{
-			Name:    name,
-			Owner:   registry.OwnerKV,
-			Timeout: item.timeout,
-			Cluster: r.MakeClusterSpec(
-				4,
-				spec.WorkloadNode(),
-				spec.RandomizeVolumeType(),
-				spec.RandomlyUseXfs(),
-			),
-			// These tests are carefully tuned to succeed up to certain number of
-			// splits; they are flaky in slower environments.
-			CompatibleClouds: registry.Clouds(spec.GCE, spec.Local),
-			Suites:           registry.Suites(registry.Nightly),
-			Leases:           item.leases,
-			Run: func(ctx context.Context, t test.Test, c cluster.Cluster) {
-				nodes := c.Spec().NodeCount - 1
-
-				settings := install.MakeClusterSettings()
-				// Enable continuous Go execution tracing to aid diagnosis of
-				// overload-driven failures. See #166402.
-				settings.ClusterSettings["obs.execution_tracer.interval"] = "10s"
-				settings.ClusterSettings["obs.execution_tracer.duration"] = "10s"
-				settings.ClusterSettings["obs.execution_tracer.total_dump_size_limit"] = "10 GiB"
-				settings.Env = append(settings.Env, "COCKROACH_MEMPROF_INTERVAL=1m", "COCKROACH_DISABLE_QUIESCENCE="+strconv.FormatBool(!item.quiesce))
-				settings.Env = append(settings.Env, item.envVars...)
-				if !item.quiesce {
-					settings.ClusterSettings["kv.raft.store_liveness.quiescence.enabled"] = "false"
-				}
-				startOpts := option.NewStartOpts(option.NoBackupSchedule)
-				startOpts.RoachprodOpts.ExtraArgs = append(startOpts.RoachprodOpts.ExtraArgs, "--cache=256MiB")
-				c.Start(ctx, t.L(), startOpts, settings, c.CRDBNodes())
-
-				t.Status("running workload")
-				workloadCtx, workloadCancel := context.WithCancel(ctx)
-				m := c.NewDeprecatedMonitor(workloadCtx, c.CRDBNodes())
-				m.Go(func(ctx context.Context) error {
-					defer workloadCancel()
-					concurrency := roachtestutil.IfLocal(c, "", " --concurrency="+fmt.Sprint(nodes*64))
-					splits := " --splits=" + roachtestutil.IfLocal(c, "2000", fmt.Sprint(item.splits))
-					cmd := fmt.Sprintf(
-						"./cockroach workload run kv --init --max-ops=1"+
-							concurrency+splits+
-							" {pgurl%s}",
-						c.CRDBNodes())
-					c.Run(ctx, option.WithNodes(c.WorkloadNode()), cmd)
-					return nil
-				})
-				m.Wait()
-			},
-		})
-	}
-}
-
 func registerKVScalability(r registry.Registry) {
 	runScalability := func(ctx context.Context, t test.Test, c cluster.Cluster, percent int) {
 		nodes := len(c.CRDBNodes())

pkg/cmd/roachtest/tests/perturbation/BUILD.bazel

@@ -14,6 +14,7 @@ go_library(
         "network_partition.go",
         "restart_node.go",
         "slow_disk.go",
+        "splits.go",
     ],
     importpath = "github.com/cockroachdb/cockroach/pkg/cmd/roachtest/tests/perturbation",
     visibility = ["//visibility:public"],
@@ -23,6 +24,7 @@ go_library(
         "//pkg/cmd/roachtest/registry",
         "//pkg/cmd/roachtest/roachtestutil",
         "//pkg/cmd/roachtest/roachtestutil/artifactsutil",
+        "//pkg/cmd/roachtest/roachtestutil/task",
         "//pkg/cmd/roachtest/spec",
         "//pkg/cmd/roachtest/test",
         "//pkg/roachprod/install",
@@ -34,6 +36,7 @@ go_library(
         "//pkg/workload/cli",
         "//pkg/workload/histogram",
         "@com_github_cockroachdb_errors//:errors",
+        "@com_github_lib_pq//:pq",
         "@com_github_stretchr_testify//require",
     ],
 )