raw_exec: oom_score_adj support (#23308)

.changelog/23308.txt

@@ -0,0 +1,3 @@
+```release-note:improvement
+raw_exec: Added support for oom_score_adj
+```

drivers/rawexec/driver.go

@@ -90,6 +90,7 @@ var (
 		"args":               hclspec.NewAttr("args", "list(string)", false),
 		"cgroup_v2_override": hclspec.NewAttr("cgroup_v2_override", "string", false),
 		"cgroup_v1_override": hclspec.NewAttr("cgroup_v1_override", "list(map(string))", false),
+		"oom_score_adj":      hclspec.NewAttr("oom_score_adj", "number", false),
 	})
 
 	// capabilities is returned by the Capabilities RPC and indicates what
@@ -156,6 +157,9 @@ type TaskConfig struct {
 	//
 	// * All resource isolation guarantees are lost FOR ALL TASKS if set *
 	OverrideCgroupV1 hclutils.MapStrStr `codec:"cgroup_v1_override"`
+
+	// OOMScoreAdj sets the oom_score_adj on Linux systems
+	OOMScoreAdj int `codec:"oom_score_adj"`
 }
 
 // TaskState is the state which is encoded in the handle returned in
@@ -324,6 +328,10 @@ func (d *Driver) StartTask(cfg *drivers.TaskConfig) (*drivers.TaskHandle, *drive
 		return nil, nil, fmt.Errorf("failed to decode driver config: %v", err)
 	}
 
+	if driverConfig.OOMScoreAdj < 0 {
+		return nil, nil, fmt.Errorf("oom_score_adj must not be negative")
+	}
+
 	d.logger.Info("starting task", "driver_cfg", hclog.Fmt("%+v", driverConfig))
 	handle := drivers.NewTaskHandle(taskHandleVersion)
 	handle.Config = cfg
@@ -353,6 +361,7 @@ func (d *Driver) StartTask(cfg *drivers.TaskConfig) (*drivers.TaskHandle, *drive
 		Resources:        cfg.Resources.Copy(),
 		OverrideCgroupV2: cgroupslib.CustomPathCG2(driverConfig.OverrideCgroupV2),
 		OverrideCgroupV1: driverConfig.OverrideCgroupV1,
+		OOMScoreAdj:      int32(driverConfig.OOMScoreAdj),
 	}
 
 	// ensure only one of cgroups_v1_override and cgroups_v2_override have been

drivers/shared/executor/executor.go

@@ -163,6 +163,10 @@ type ExecCommand struct {
 	//
 	// * All resource isolation guarantees are lost FOR ALL TASKS if set *
 	OverrideCgroupV1 map[string]string
+
+	// OOMScoreAdj allows setting oom_score_adj (likelihood of process being
+	// OOM killed) on Linux systems
+	OOMScoreAdj int32
 }
 
 func (c *ExecCommand) getCgroupOr(controller, fallback string) string {

drivers/shared/executor/executor_universal_linux.go

@@ -121,8 +121,8 @@ func (e *UniversalExecutor) statCG(cgroup string) (int, func(), error) {
 func (e *UniversalExecutor) configureResourceContainer(command *ExecCommand, pid int) (func(), error) {
 	cgroup := command.StatsCgroup()
 
-	// ensure tasks do not inherit Nomad agent oom_score_adj value
-	if err := e.setOomAdj(); err != nil {
+	// ensure tasks get the desired oom_score_adj value set
+	if err := e.setOomAdj(command.OOMScoreAdj); err != nil {
 		return nil, err
 	}
 
@@ -280,12 +280,10 @@ func (e *UniversalExecutor) configureCG2(cgroup string, command *ExecCommand) {
 	_ = ed.Write("cpuset.cpus", cpusetCpus)
 }
 
-func (e *UniversalExecutor) setOomAdj() error {
-	// children should not inherit Nomad agent oom_score_adj value
-	//
+func (e *UniversalExecutor) setOomAdj(oomScore int32) error {
 	// /proc/self/oom_score_adj should work on both cgroups v1 and v2 systems
 	// range is -1000 to 1000; 0 is the default
-	return os.WriteFile("/proc/self/oom_score_adj", []byte("0"), 0644)
+	return os.WriteFile("/proc/self/oom_score_adj", []byte(strconv.Itoa(int(oomScore))), 0644)
 }
 
 func (*UniversalExecutor) computeCPU(command *ExecCommand) uint64 {

drivers/shared/executor/executor_universal_linux_test.go

@@ -7,6 +7,9 @@ package executor
 
 import (
 	"fmt"
+	"os"
+	"strconv"
+	"strings"
 	"testing"
 
 	"github.com/hashicorp/nomad/ci"
@@ -99,3 +102,28 @@ func TestExecutor_InvalidCgroup(t *testing.T) {
 	must.ErrorContains(t, err, "unable to configure cgroups: no such file or directory")
 
 }
+
+func TestUniversalExecutor_setOomAdj(t *testing.T) {
+	ci.Parallel(t)
+
+	factory := universalFactory
+	testExecCmd := testExecutorCommand(t)
+	execCmd, allocDir := testExecCmd.command, testExecCmd.allocDir
+	execCmd.Cmd = "sleep"
+	execCmd.Args = []string{"infinity"}
+	execCmd.OOMScoreAdj = 1000
+
+	factory.configureExecCmd(t, execCmd)
+	defer allocDir.Destroy()
+	executor := factory.new(testlog.HCLogger(t), compute)
+	defer executor.Shutdown("", 0)
+
+	p, err := executor.Launch(execCmd)
+	must.NoError(t, err)
+
+	oomScore, err := os.ReadFile(fmt.Sprintf("/proc/%d/oom_score_adj", p.Pid))
+	must.NoError(t, err)
+
+	oomScoreInt, _ := strconv.Atoi(strings.TrimSuffix(string(oomScore), "\n"))
+	must.Eq(t, execCmd.OOMScoreAdj, int32(oomScoreInt))
+}

drivers/shared/executor/grpc_client.go

@@ -54,6 +54,7 @@ func (c *grpcExecutorClient) Launch(cmd *ExecCommand) (*ProcessState, error) {
 		Capabilities:     cmd.Capabilities,
 		CgroupV2Override: cmd.OverrideCgroupV2,
 		CgroupV1Override: cmd.OverrideCgroupV1,
+		OomScoreAdj:      cmd.OOMScoreAdj,
 	}
 	resp, err := c.client.Launch(ctx, req)
 	if err != nil {

drivers/shared/executor/grpc_server.go

@@ -43,6 +43,7 @@ func (s *grpcExecutorServer) Launch(ctx context.Context, req *proto.LaunchReques
 		Capabilities:     req.Capabilities,
 		OverrideCgroupV2: req.CgroupV2Override,
 		OverrideCgroupV1: req.CgroupV1Override,
+		OOMScoreAdj:      req.OomScoreAdj,
 	})
 
 	if err != nil {