Add profiling to dockerutil

Adds profiling with `runsc debug` or pprof to dockerutil. All
targets using dockerutil should now be able to use profiling.

In addition, modifies existing benchmarks to use profiling.

PiperOrigin-RevId: 323298634

8 files changed, 126 insertions, 56 deletions

diff --git a/test/benchmarks/README.md b/test/benchmarks/README.md
index 9ff602cf1..d1bbabf6f 100644
--- a/test/benchmarks/README.md
+++ b/test/benchmarks/README.md
@@ -13,33 +13,51 @@ To run benchmarks you will need:
 
 *   Docker installed (17.09.0 or greater).
 
-The easiest way to run benchmarks is to use the script at
-//scripts/benchmark.sh.
+The easiest way to setup runsc for running benchmarks is to use the make file.
+From the root directory:
 
-If not using the script, you will need:
+*   Download images: `make load-all-images`
+*   Install runsc suitable for benchmarking, which should probably not have
+    strace or debug logs enabled. For example:`make configure RUNTIME=myrunsc
+    ARGS=--platform=kvm`.
+*   Restart docker: `sudo service docker restart`
 
-*   `runsc` configured with docker
+You should now have a runtime with the following options configured in
+`/etc/docker/daemon.json`
 
-Note: benchmarks call the runtime by name. If docker can run it with
-`--runtime=` flag, these tools should work.
+```
+"myrunsc": {
+            "path": "/tmp/myrunsc/runsc",
+            "runtimeArgs": [
+                "--debug-log",
+                "/tmp/bench/logs/runsc.log.%TEST%.%TIMESTAMP%.%COMMAND%",
+                "--platform=kvm"
+            ]
+        },
+
+```
+
+This runtime has been configured with a debugging off and strace logs off and is
+using kvm for demonstration.
 
 ## Running benchmarks
 
-The easiest way to run is with the script at //scripts/benchmarks.sh. The script
-will run all benchmarks under //test/benchmarks if a target is not provided.
+Given the runtime above runtime `myrunsc`, run benchmarks with the following:
 
-```bash
-./script/benchmarks.sh //path/to/target
+```
+make sudo TARGETS=//path/to:target ARGS="--runtime=myrunsc -test.v \
+  -test.bench=." OPTIONS="-c opt
 ```
 
-If you want to run benchmarks manually:
-
-*   Run `make load-all-images` from `//`
-*   Run with:
+For example, to run only the Iperf tests:
 
-```bash
-bazel test --test_arg=--runtime=RUNTIME -c opt --test_output=streamed --test_timeout=600 --test_arg=-test.bench=. --nocache_test_results //path/to/target
 ```
+make sudo TARGETS=//test/benchmarks/network:network_test \
+  ARGS="--runtime=myrunsc -test.v -test.bench=Iperf" OPTIONS="-c opt"
+```
+
+Benchmarks are run with root as some benchmarks require root privileges to do
+things like drop caches.
 
 ## Writing benchmarks
 
@@ -69,6 +87,7 @@ var h harness.Harness
 func BenchmarkMyCoolOne(b *testing.B) {
   machine, err := h.GetMachine()
   // check err
+  defer machine.CleanUp()
 
   ctx := context.Background()
   container := machine.GetContainer(ctx, b)
@@ -82,7 +101,7 @@ func BenchmarkMyCoolOne(b *testing.B) {
       Image: "benchmarks/my-cool-image",
       Env: []string{"MY_VAR=awesome"},
       other options...see dockerutil
-    }, "sh", "-c", "echo MY_VAR" ...)
+    }, "sh", "-c", "echo MY_VAR")
     //check err
     b.StopTimer()
 
@@ -107,12 +126,32 @@ Some notes on the above:
     flags, remote virtual machines (eventually), and other services.
 *   Respect `b.N` in that users of the benchmark may want to "run for an hour"
     or something of the sort.
-*   Use the `b.ReportMetric` method to report custom metrics.
+*   Use the `b.ReportMetric()` method to report custom metrics.
 *   Set the timer if time is useful for reporting. There isn't a way to turn off
     default metrics in testing.B (B/op, allocs/op, ns/op).
 *   Take a look at dockerutil at //pkg/test/dockerutil to see all methods
     available from containers. The API is based on the "official"
     [docker API for golang](https://pkg.go.dev/mod/github.com/docker/docker).
-*   `harness.GetMachine` marks how many machines this tests needs. If you have a
-    client and server and to mark them as multiple machines, call it
-    `GetMachine` twice.
+*   `harness.GetMachine()` marks how many machines this tests needs. If you have
+    a client and server and to mark them as multiple machines, call
+    `harness.GetMachine()` twice.
+
+## Profiling
+
+For profiling, the runtime is required to have the `--profile` flag enabled.
+This flag loosens seccomp filters so that the runtime can write profile data to
+disk. This configuration is not recommended for production.
+
+*   Install runsc with the `--profile` flag: `make configure RUNTIME=myrunsc
+    ARGS="--profile --platform=kvm --vfs2"`. The kvm and vfs2 flags are not
+    required, but are included for demonstration.
+*   Restart docker: `sudo service docker restart`
+
+To run and generate CPU profiles fs_test test run:
+
+```
+make sudo TARGETS=//test/benchmarks/fs:fs_test \
+  ARGS="--runtime=myrunsc -test.v -test.bench=. --pprof-cpu" OPTIONS="-c opt"
+```
+
+Profiles would be at: `/tmp/profile/myrunsc/CONTAINERNAME/cpu.pprof`
diff --git a/test/benchmarks/fs/bazel_test.go b/test/benchmarks/fs/bazel_test.go
index fdcac1a7a..9b652fd43 100644
--- a/test/benchmarks/fs/bazel_test.go
+++ b/test/benchmarks/fs/bazel_test.go
@@ -15,6 +15,7 @@ package fs
 
 import (
 	"context"
+	"fmt"
 	"strings"
 	"testing"
 
@@ -51,10 +52,10 @@ func BenchmarkABSL(b *testing.B) {
 
 			workdir := "/abseil-cpp"
 
-			// Start a container.
+			// Start a container and sleep by an order of b.N.
 			if err := container.Spawn(ctx, dockerutil.RunOpts{
 				Image: "benchmarks/absl",
-			}, "sleep", "1000"); err != nil {
+			}, "sleep", fmt.Sprintf("%d", 1000000)); err != nil {
 				b.Fatalf("run failed with: %v", err)
 			}
 
@@ -67,15 +68,21 @@ func BenchmarkABSL(b *testing.B) {
 				workdir = "/tmp" + workdir
 			}
 
-			// Drop Caches.
-			if bm.clearCache {
-				if out, err := machine.RunCommand("/bin/sh -c sync; echo 3 > /proc/sys/vm/drop_caches"); err != nil {
-					b.Fatalf("failed to drop caches: %v %s", err, out)
-				}
-			}
-
+			// Restart profiles after the copy.
+			container.RestartProfiles()
 			b.ResetTimer()
+			// Drop Caches and bazel clean should happen inside the loop as we may use
+			// time options with b.N. (e.g. Run for an hour.)
 			for i := 0; i < b.N; i++ {
+				b.StopTimer()
+				// Drop Caches for clear cache runs.
+				if bm.clearCache {
+					if out, err := machine.RunCommand("/bin/sh", "-c", "sync && sysctl vm.drop_caches=3"); err != nil {
+						b.Skipf("failed to drop caches: %v %s. You probably need root.", err, out)
+					}
+				}
+				b.StartTimer()
+
 				got, err := container.Exec(ctx, dockerutil.ExecOpts{
 					WorkDir: workdir,
 				}, "bazel", "build", "-c", "opt", "absl/base/...")
@@ -88,6 +95,13 @@ func BenchmarkABSL(b *testing.B) {
 				if !strings.Contains(got, want) {
 					b.Fatalf("string %s not in: %s", want, got)
 				}
+				// Clean bazel in case we use b.N.
+				_, err = container.Exec(ctx, dockerutil.ExecOpts{
+					WorkDir: workdir,
+				}, "bazel", "clean")
+				if err != nil {
+					b.Fatalf("build failed with: %v", err)
+				}
 				b.StartTimer()
 			}
 		})
diff --git a/test/benchmarks/harness/machine.go b/test/benchmarks/harness/machine.go
index 93c0db9ce..88e5e841b 100644
--- a/test/benchmarks/harness/machine.go
+++ b/test/benchmarks/harness/machine.go
@@ -25,9 +25,14 @@ import (
 
 // Machine describes a real machine for use in benchmarks.
 type Machine interface {
-	// GetContainer gets a container from the machine,
+	// GetContainer gets a container from the machine. The container uses the
+	// runtime under test and is profiled if requested by flags.
 	GetContainer(ctx context.Context, log testutil.Logger) *dockerutil.Container
 
+	// GetNativeContainer gets a native container from the machine. Native containers
+	// use runc by default and are not profiled.
+	GetNativeContainer(ctx context.Context, log testutil.Logger) *dockerutil.Container
+
 	// RunCommand runs cmd on this machine.
 	RunCommand(cmd string, args ...string) (string, error)
 
@@ -47,6 +52,11 @@ func (l *localMachine) GetContainer(ctx context.Context, logger testutil.Logger)
 	return dockerutil.MakeContainer(ctx, logger)
 }
 
+// GetContainer implements Machine.GetContainer for localMachine.
+func (l *localMachine) GetNativeContainer(ctx context.Context, logger testutil.Logger) *dockerutil.Container {
+	return dockerutil.MakeNativeContainer(ctx, logger)
+}
+
 // RunCommand implements Machine.RunCommand for localMachine.
 func (l *localMachine) RunCommand(cmd string, args ...string) (string, error) {
 	c := exec.Command(cmd, args...)
diff --git a/test/benchmarks/harness/util.go b/test/benchmarks/harness/util.go
index cc7de6426..7f8e42201 100644
--- a/test/benchmarks/harness/util.go
+++ b/test/benchmarks/harness/util.go
@@ -27,7 +27,7 @@ import (
 // IP:port.
 func WaitUntilServing(ctx context.Context, machine Machine, server net.IP, port int) error {
 	var logger testutil.DefaultLogger = "netcat"
-	netcat := machine.GetContainer(ctx, logger)
+	netcat := machine.GetNativeContainer(ctx, logger)
 	defer netcat.CleanUp(ctx)
 
 	cmd := fmt.Sprintf("while ! nc -zv %s %d; do true; done", server.String(), port)
diff --git a/test/benchmarks/network/BUILD b/test/benchmarks/network/BUILD
index 16d267bc8..363041fb7 100644
--- a/test/benchmarks/network/BUILD
+++ b/test/benchmarks/network/BUILD
@@ -24,6 +24,7 @@ go_test(
     ],
     deps = [
         "//pkg/test/dockerutil",
+        "//pkg/test/testutil",
         "//test/benchmarks/harness",
     ],
 )
diff --git a/test/benchmarks/network/httpd_test.go b/test/benchmarks/network/httpd_test.go
index f9afdf15f..fe23ca949 100644
--- a/test/benchmarks/network/httpd_test.go
+++ b/test/benchmarks/network/httpd_test.go
@@ -52,12 +52,12 @@ func BenchmarkHttpdConcurrency(b *testing.B) {
 	defer serverMachine.CleanUp()
 
 	// The test iterates over client concurrency, so set other parameters.
-	requests := 1000
+	requests := 10000
 	concurrency := []int{1, 5, 10, 25}
 	doc := docs["10Kb"]
 
 	for _, c := range concurrency {
-		b.Run(fmt.Sprintf("%dConcurrency", c), func(b *testing.B) {
+		b.Run(fmt.Sprintf("%d", c), func(b *testing.B) {
 			runHttpd(b, clientMachine, serverMachine, doc, requests, c)
 		})
 	}
@@ -78,7 +78,7 @@ func BenchmarkHttpdDocSize(b *testing.B) {
 	}
 	defer serverMachine.CleanUp()
 
-	requests := 1000
+	requests := 10000
 	concurrency := 1
 
 	for name, filename := range docs {
@@ -129,7 +129,7 @@ func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc st
 	harness.WaitUntilServing(ctx, clientMachine, ip, servingPort)
 
 	// Grab a client.
-	client := clientMachine.GetContainer(ctx, b)
+	client := clientMachine.GetNativeContainer(ctx, b)
 	defer client.CleanUp(ctx)
 
 	path := fmt.Sprintf("http://%s:%d/%s", ip, servingPort, doc)
@@ -137,6 +137,7 @@ func runHttpd(b *testing.B, clientMachine, serverMachine harness.Machine, doc st
 	cmd = fmt.Sprintf("ab -n %d -c %d %s", requests, concurrency, path)
 
 	b.ResetTimer()
+	server.RestartProfiles()
 	for i := 0; i < b.N; i++ {
 		out, err := client.Run(ctx, dockerutil.RunOpts{
 			Image: "benchmarks/ab",
diff --git a/test/benchmarks/network/iperf_test.go b/test/benchmarks/network/iperf_test.go
index 664e0797e..a5e198e14 100644
--- a/test/benchmarks/network/iperf_test.go
+++ b/test/benchmarks/network/iperf_test.go
@@ -22,12 +22,13 @@ import (
 	"testing"
 
 	"gvisor.dev/gvisor/pkg/test/dockerutil"
+	"gvisor.dev/gvisor/pkg/test/testutil"
 	"gvisor.dev/gvisor/test/benchmarks/harness"
 )
 
 func BenchmarkIperf(b *testing.B) {
+	const time = 10 // time in seconds to run the client.
 
-	// Get two machines
 	clientMachine, err := h.GetMachine()
 	if err != nil {
 		b.Fatalf("failed to get machine: %v", err)
@@ -39,30 +40,32 @@ func BenchmarkIperf(b *testing.B) {
 		b.Fatalf("failed to get machine: %v", err)
 	}
 	defer serverMachine.CleanUp()
-
+	ctx := context.Background()
 	for _, bm := range []struct {
-		name          string
-		clientRuntime string
-		serverRuntime string
+		name       string
+		clientFunc func(context.Context, testutil.Logger) *dockerutil.Container
+		serverFunc func(context.Context, testutil.Logger) *dockerutil.Container
 	}{
 		// We are either measuring the server or the client. The other should be
 		// runc. e.g. Upload sees how fast the runtime under test uploads to a native
 		// server.
-		{name: "Upload", clientRuntime: dockerutil.Runtime(), serverRuntime: "runc"},
-		{name: "Download", clientRuntime: "runc", serverRuntime: dockerutil.Runtime()},
+		{
+			name:       "Upload",
+			clientFunc: clientMachine.GetContainer,
+			serverFunc: serverMachine.GetNativeContainer,
+		},
+		{
+			name:       "Download",
+			clientFunc: clientMachine.GetNativeContainer,
+			serverFunc: serverMachine.GetContainer,
+		},
 	} {
 		b.Run(bm.name, func(b *testing.B) {
-
-			// Get a container from the server and set its runtime.
-			ctx := context.Background()
-			server := serverMachine.GetContainer(ctx, b)
+			// Set up the containers.
+			server := bm.serverFunc(ctx, b)
 			defer server.CleanUp(ctx)
-			server.Runtime = bm.serverRuntime
-
-			// Get a container from the client and set its runtime.
-			client := clientMachine.GetContainer(ctx, b)
+			client := bm.clientFunc(ctx, b)
 			defer client.CleanUp(ctx)
-			client.Runtime = bm.clientRuntime
 
 			// iperf serves on port 5001 by default.
 			port := 5001
@@ -91,11 +94,14 @@ func BenchmarkIperf(b *testing.B) {
 			}
 
 			// iperf report in Kb realtime
-			cmd := fmt.Sprintf("iperf -f K --realtime -c %s -p %d", ip.String(), servingPort)
+			cmd := fmt.Sprintf("iperf -f K --realtime --time %d -c %s -p %d", time, ip.String(), servingPort)
 
 			// Run the client.
 			b.ResetTimer()
 
+			// Restart the server profiles. If the server isn't being profiled
+			// this does nothing.
+			server.RestartProfiles()
 			for i := 0; i < b.N; i++ {
 				out, err := client.Run(ctx, dockerutil.RunOpts{
 					Image: "benchmarks/iperf",
diff --git a/test/packetimpact/runner/packetimpact_test.go b/test/packetimpact/runner/packetimpact_test.go
index 1a0221893..74e1e6def 100644
--- a/test/packetimpact/runner/packetimpact_test.go
+++ b/test/packetimpact/runner/packetimpact_test.go
@@ -142,7 +142,7 @@ func TestOne(t *testing.T) {
 	// Create the Docker container for the DUT.
 	dut := dockerutil.MakeContainer(ctx, logger("dut"))
 	if *dutPlatform == "linux" {
-		dut.Runtime = ""
+		dut = dockerutil.MakeNativeContainer(ctx, logger("dut"))
 	}
 
 	runOpts := dockerutil.RunOpts{
@@ -208,8 +208,7 @@ func TestOne(t *testing.T) {
 	}
 
 	// Create the Docker container for the testbench.
-	testbench := dockerutil.MakeContainer(ctx, logger("testbench"))
-	testbench.Runtime = "" // The testbench always runs on Linux.
+	testbench := dockerutil.MakeNativeContainer(ctx, logger("testbench"))
 
 	tbb := path.Base(*testbenchBinary)
 	containerTestbenchBinary := "/packetimpact/" + tbb