feat(envd): give envd realtime IO priority, reset for user processes (#2681)

envd at IOSchedulingClass=realtime + IOWeight=10000; lower io.weight on
user/PTY/socat sub-cgroups; user-spawned processes get ioprio reset via
the existing wrapper.

Author: ValentaTomas

packages/envd/internal/services/cgroups/cgroup2.go

@@ -104,15 +104,34 @@ func createCgroups(configs cgroup2Config) (map[ProcessType]int, error) {
 	return results, nil
 }
 
+// writeCgroupProp writes a cgroupfs property without O_CREATE so missing
+// properties error out rather than being silently created on a tmpfs fallback.
+func writeCgroupProp(path, value string) error {
+	f, err := os.OpenFile(path, os.O_WRONLY|os.O_TRUNC, 0)
+	if err != nil {
+		return err
+	}
+	defer f.Close()
+	_, err = f.WriteString(value)
+
+	return err
+}
+
 func createCgroup(fullPath string, properties map[string]string) (int, error) {
 	if err := os.MkdirAll(fullPath, 0o755); err != nil {
 		return -1, fmt.Errorf("failed to create cgroup root: %w", err)
 	}
 
 	var errs []error
 	for name, value := range properties {
-		if err := os.WriteFile(filepath.Join(fullPath, name), []byte(value), 0o644); err != nil {
-			errs = append(errs, fmt.Errorf("failed to write cgroup property: %w", err))
+		if err := writeCgroupProp(filepath.Join(fullPath, name), value); err != nil {
+			// Skip properties whose controller isn't enabled in subtree_control.
+			if errors.Is(err, os.ErrNotExist) || errors.Is(err, os.ErrPermission) {
+				fmt.Fprintf(os.Stderr, "cgroup property %q unavailable at %q, skipping\n", name, fullPath)
+
+				continue
+			}
+			errs = append(errs, fmt.Errorf("failed to write cgroup property %q: %w", name, err))
 		}
 	}
 	if len(errs) > 0 {

packages/envd/internal/services/process/handler/handler.go

@@ -96,12 +96,9 @@ func New(
 	// User command string for logging (without the internal wrapper details).
 	userCmd := strings.Join(append([]string{req.GetProcess().GetCmd()}, req.GetProcess().GetArgs()...), " ")
 
-	// Wrap the command in a shell that sets the OOM score and nice value before exec-ing the actual command.
-	// This eliminates the race window where grandchildren could inherit the parent's protected OOM score (-1000)
-	// or high CPU priority (nice -20) before the post-start calls had a chance to correct them.
-	// nice(1) applies a relative adjustment, so we compute the delta from the current (inherited) nice to the target.
+	// Wrap in a shell that resets oom_score_adj, ioprio (ionice best-effort/4), and nice.
 	niceDelta := defaultNice - currentNice()
-	oomWrapperScript := fmt.Sprintf(`echo %d > /proc/$$/oom_score_adj && exec /usr/bin/nice -n %d "${@}"`, defaultOomScore, niceDelta)
+	oomWrapperScript := fmt.Sprintf(`echo %d > /proc/$$/oom_score_adj && exec /usr/bin/ionice -c 2 -n 4 /usr/bin/nice -n %d "${@}"`, defaultOomScore, niceDelta)
 	wrapperArgs := append([]string{"-c", oomWrapperScript, "--", req.GetProcess().GetCmd()}, req.GetProcess().GetArgs()...)
 	cmd := exec.CommandContext(ctx, "/bin/sh", wrapperArgs...)
 

packages/envd/main.go

@@ -242,19 +242,22 @@ func createCgroupManager() (m cgroups.Manager) {
 
 	opts := []cgroups.Cgroup2ManagerOption{
 		cgroups.WithCgroup2ProcessType(cgroups.ProcessTypePTY, "ptys", map[string]string{
-			"cpu.weight":  "200", // gets much preferred cpu access, to help keep these real time
+			"cpu.weight":  "200",
+			"io.weight":   "default 50",
 			"memory.high": fmt.Sprintf("%d", memoryHigh),
 			"memory.max":  fmt.Sprintf("%d", memoryMax),
 		}),
 		cgroups.WithCgroup2ProcessType(cgroups.ProcessTypeSocat, "socats", map[string]string{
-			"cpu.weight": "150", // gets slightly preferred cpu access
+			"cpu.weight": "150",
+			"io.weight":  "default 50",
 			"memory.min": fmt.Sprintf("%d", 5*megabyte),
 			"memory.low": fmt.Sprintf("%d", 8*megabyte),
 		}),
 		cgroups.WithCgroup2ProcessType(cgroups.ProcessTypeUser, "user", map[string]string{
 			"memory.high": fmt.Sprintf("%d", memoryHigh),
 			"memory.max":  fmt.Sprintf("%d", memoryMax),
-			"cpu.weight":  "50", // less than envd, and less than core processes that default to 100
+			"cpu.weight":  "50",
+			"io.weight":   "default 10",
 		}),
 	}
 	if cgroupRoot != "" {

packages/orchestrator/pkg/template/build/core/rootfs/files/envd.service.tpl

@@ -17,6 +17,8 @@ LimitCORE=infinity
 ExecStartPre=/bin/sh -c 'mountpoint -q /etc/ssl/certs || (mkdir -p /run/e2b/certs && mount --bind /run/e2b/certs /etc/ssl/certs) && ([ -s /etc/ssl/certs/ca-certificates.crt ] || update-ca-certificates)'
 ExecStart=/bin/bash -l -c "/usr/bin/envd"
 Nice=-20
+IOSchedulingClass=realtime
+IOSchedulingPriority=4
 OOMPolicy=continue
 OOMScoreAdjust=-1000
 Environment="GOMEMLIMIT={{ .MemoryLimit }}MiB"
@@ -26,6 +28,8 @@ MemoryMin=50M
 MemoryLow=100M
 CPUAccounting=yes
 CPUWeight=1000
+IOAccounting=yes
+IOWeight=10000
 
 [Install]
 WantedBy=multi-user.target