#15 control swap and memory reservation

README.md

@@ -23,6 +23,7 @@ Contributions are welcome, of course.
 * Container log is forwarded to [Nomad logger](https://www.nomadproject.io/docs/commands/alloc/logs.html) 
 * utilize podmans --init feature
 * set username or UID used for the specified command within the container (podman --user option).
+* fine tune memory usage: standard [nomad memory resource](https://www.nomadproject.io/docs/job-specification/resources.html#memory) plus additional driver specific swap, swappiness and reservation parameters, OOM handling
 
 
 ## Building The Driver from source
@@ -152,6 +153,34 @@ config {
 
 ```
 
+* **memory_reservation** - Memory soft limit (nit = b (bytes), k (kilobytes), m (megabytes), or g (gigabytes))
+
+After setting memory reservation, when the system detects memory contention or low memory, containers are forced to restrict their consumption to their reservation. So you should always set the value below --memory, otherwise the hard limit will take precedence. By default, memory reservation will be the same as memory limit.
+
+```
+config {
+  memory_reservation = "100m"
+}
+```
+
+* **memory_swap** - A limit value equal to memory plus swap. The swap LIMIT should always be larger than the [memory value](https://www.nomadproject.io/docs/job-specification/resources.html#memory). 
+
+Unit can be b (bytes), k (kilobytes), m (megabytes), or g (gigabytes). If you don't specify a unit, b is used. Set LIMIT to -1 to enable unlimited swap.
+
+```
+config {
+  memory_swap = "180m"
+}
+```
+
+* **memory_swappiness** - Tune a container's memory swappiness behavior. Accepts an integer between 0 and 100.
+
+```
+config {
+  memory_swap = "180m"
+}
+```
+
 ## Example job
 
 ```

config.go

@@ -48,14 +48,17 @@ var (
 	// taskConfigSpec is the hcl specification for the driver config section of
 	// a task within a job. It is returned in the TaskConfigSchema RPC
 	taskConfigSpec = hclspec.NewObject(map[string]*hclspec.Spec{
-		"image":     hclspec.NewAttr("image", "string", true),
-		"command":   hclspec.NewAttr("command", "string", false),
-		"hostname":  hclspec.NewAttr("hostname", "string", false),
-		"init":      hclspec.NewAttr("init", "bool", false),
-		"init_path": hclspec.NewAttr("init_path", "string", false),
-		"args":      hclspec.NewAttr("args", "list(string)", false),
-		"volumes":   hclspec.NewAttr("volumes", "list(string)", false),
-		"port_map":  hclspec.NewAttr("port_map", "list(map(number))", false),
+		"image":              hclspec.NewAttr("image", "string", true),
+		"command":            hclspec.NewAttr("command", "string", false),
+		"hostname":           hclspec.NewAttr("hostname", "string", false),
+		"init":               hclspec.NewAttr("init", "bool", false),
+		"init_path":          hclspec.NewAttr("init_path", "string", false),
+		"args":               hclspec.NewAttr("args", "list(string)", false),
+		"volumes":            hclspec.NewAttr("volumes", "list(string)", false),
+		"port_map":           hclspec.NewAttr("port_map", "list(map(number))", false),
+		"memory_reservation": hclspec.NewAttr("memory_reservation", "string", false),
+		"memory_swap":        hclspec.NewAttr("memory_swap", "string", false),
+		"memory_swappiness":  hclspec.NewAttr("memory_swappiness", "number", false),
 	})
 )
 
@@ -78,12 +81,15 @@ type PluginConfig struct {
 
 // TaskConfig is the driver configuration of a task within a job
 type TaskConfig struct {
-	Image    string             `codec:"image"`
-	Command  string             `codec:"command"`
-	Args     []string           `codec:"args"`
-	Volumes  []string           `codec:"volumes"`
-	Hostname string             `codec:"hostname"`
-	Init     bool               `codec:"init"`
-	InitPath string             `codec:"init_path"`
-	PortMap  hclutils.MapStrInt `codec:"port_map"`
+	Image             string             `codec:"image"`
+	Command           string             `codec:"command"`
+	Args              []string           `codec:"args"`
+	Volumes           []string           `codec:"volumes"`
+	Hostname          string             `codec:"hostname"`
+	Init              bool               `codec:"init"`
+	InitPath          string             `codec:"init_path"`
+	PortMap           hclutils.MapStrInt `codec:"port_map"`
+	MemoryReservation string             `codec:"memory_reservation"`
+	MemorySwap        string             `codec:"memory_swap"`
+	MemorySwappiness  int64              `codec:"memory_swappiness"`
 }

driver.go

@@ -321,19 +321,26 @@ func (d *Driver) StartTask(cfg *drivers.TaskConfig) (*drivers.TaskHandle, *drive
 		}
 	}
 
+	swap := memoryLimit
+	if driverConfig.MemorySwap != "" {
+		swap = driverConfig.MemorySwap
+	}
+
 	createOpts := iopodman.Create{
-		Args:       allArgs,
-		Env:        &allEnv,
-		Name:       &containerName,
-		Volume:     &allVolumes,
-		Memory:     &memoryLimit,
-		MemorySwap: &memoryLimit,
-		CpuShares:  &cpuShares,
-		LogOpt:     &logOpts,
-		Hostname:   &driverConfig.Hostname,
-		Init:       &driverConfig.Init,
-		InitPath:   &driverConfig.InitPath,
-		User:		&cfg.User,
+		Args:              allArgs,
+		Env:               &allEnv,
+		Name:              &containerName,
+		Volume:            &allVolumes,
+		Memory:            &memoryLimit,
+		CpuShares:         &cpuShares,
+		LogOpt:            &logOpts,
+		Hostname:          &driverConfig.Hostname,
+		Init:              &driverConfig.Init,
+		InitPath:          &driverConfig.InitPath,
+		User:              &cfg.User,
+		MemoryReservation: &driverConfig.MemoryReservation,
+		MemorySwap:        &swap,
+		MemorySwappiness:  &driverConfig.MemorySwappiness,
 	}
 
 	// Setup port mapping and exposed ports

driver_test.go

@@ -639,7 +639,7 @@ func TestPodmanDriver_OOM(t *testing.T) {
 	taskCfg := newTaskConfig("", []string{
 		// Incrementally creates a bigger and bigger variable.
 		"sh",
-		"-c", 
+		"-c",
 		"x=a; while true; do eval x='$x$x'; done",
 	})
 	// enable --init
@@ -726,6 +726,62 @@ func TestPodmanDriver_User(t *testing.T) {
 
 }
 
+// test memory/swap options
+func TestPodmanDriver_Swap(t *testing.T) {
+	if !tu.IsCI() {
+		t.Parallel()
+	}
+
+	taskCfg := newTaskConfig("", busyboxLongRunningCmd)
+	task := &drivers.TaskConfig{
+		ID:        uuid.Generate(),
+		Name:      "swap",
+		AllocID:   uuid.Generate(),
+		Resources: createBasicResources(),
+	}
+	// limit memory to 50MB
+	task.Resources.NomadResources.Memory.MemoryMB = 50
+	// but reserve 40MB
+	taskCfg.MemoryReservation = "40m"
+	// and allow mem+swap of 100MB (= 50 MB Swap)
+	taskCfg.MemorySwap = "100m"
+	// set a swappiness of 60
+	taskCfg.MemorySwappiness = 60
+	require.NoError(t, task.EncodeConcreteDriverConfig(&taskCfg))
+
+	containerName := BuildContainerName(task)
+
+	d := podmanDriverHarness(t, nil)
+	cleanup := d.MkAllocDir(task, true)
+	defer cleanup()
+
+	_, _, err := d.StartTask(task)
+	require.NoError(t, err)
+
+	defer d.DestroyTask(task.ID, true)
+
+	// Attempt to wait
+	waitCh, err := d.WaitTask(context.Background(), task.ID)
+	require.NoError(t, err)
+
+	select {
+	case <-waitCh:
+		t.Fatalf("wait channel should not have received an exit result")
+	case <-time.After(time.Duration(tu.TestMultiplier()*1) * time.Second):
+	}
+	// inspect container to learn about the actual podman limits
+	inspectJSON := inspectContainer(t, containerName)
+	var inspectData iopodman.InspectContainerData
+
+	require.NoError(t, json.Unmarshal([]byte(inspectJSON), &inspectData))
+	// see if the configured values are set correctly
+	require.Equal(t, int64(52428800), inspectData.HostConfig.Memory)
+	require.Equal(t, int64(41943040), inspectData.HostConfig.MemoryReservation)
+	require.Equal(t, int64(104857600), inspectData.HostConfig.MemorySwap)
+	require.Equal(t, int64(60), inspectData.HostConfig.MemorySwappiness)
+
+}
+
 // read a tasks logfile into a string, fail on error
 func readLogfile(t *testing.T, task *drivers.TaskConfig) string {
 	logfile := filepath.Join(filepath.Dir(task.StdoutPath), fmt.Sprintf("%s.stdout.0", task.Name))

iopodman/inspect.go

@@ -220,4 +220,19 @@ type InspectContainerHostConfig struct {
 	// and represents the container port. A single container port may be
 	// bound to multiple host ports (on different IPs).
 	PortBindings map[string][]InspectHostPort `json:"PortBindings"`
+	// Memory indicates the memory resources allocated to the container.
+	// This is the limit (in bytes) of RAM the container may use.
+	Memory int64 `json:"Memory"`
+	// MemoryReservation is the reservation (soft limit) of memory available
+	// to the container. Soft limits are warnings only and can be exceeded.
+	MemoryReservation int64 `json:"MemoryReservation"`
+	// MemorySwap is the total limit for all memory available to the
+	// container, including swap. 0 indicates that there is no limit to the
+	// amount of memory available.
+	MemorySwap int64 `json:"MemorySwap"`
+	// MemorySwappiness is the willingness of the kernel to page container
+	// memory to swap. It is an integer from 0 to 100, with low numbers
+	// being more likely to be put into swap.
+	// -1, the default, will not set swappiness and use the system defaults.
+	MemorySwappiness int64 `json:"MemorySwappiness"`
 }