Merge branch 'main' into fix/external-trace-id-per-run

Author: nicktrn

.server-changes/ttl-runs-no-longer-stuck-after-requeue.md

@@ -0,0 +1,6 @@
+---
+area: webapp
+type: fix
+---
+
+Runs triggered with a `ttl` could get permanently stuck in the queued state if they started executing and were then requeued after a failure (for example a worker dying mid-run) once the TTL had already elapsed. Requeued runs now dequeue normally: a run's TTL only applies while it is waiting to start for the first time.

apps/supervisor/src/env.ts

@@ -220,6 +220,8 @@ export const Env = z
     KUBERNETES_RUNNER_SECCOMP_PROFILE_RUNTIMES: z
       .enum(["none", "node-24-plus", "all"])
       .default("node-24-plus"),
+    KUBERNETES_RUNNER_SECURITY_CONTEXT: z.enum(["off", "baseline", "restricted"]).default("off"),
+    KUBERNETES_RUNNER_RUN_AS_USER: z.coerce.number().int().min(1).default(1000),
 
     // Pod DNS config — override the cluster default ndots to `KUBERNETES_POD_DNS_NDOTS`.
     // Default k8s ndots is 5: any name with fewer than 5 dots (e.g. `api.example.com`, 2 dots) is first walked

apps/supervisor/src/workloadManager/kubernetes.test.ts

@@ -2,6 +2,7 @@ import { describe, expect, it } from "vitest";
 import {
   nodetypeNodeSelector,
   runPodTolerations,
+  runnerSecurityContext,
   withRunnerSeccompProfile,
   withNodeSelector,
 } from "./kubernetesPodSpec.js";
@@ -153,3 +154,35 @@ describe("withRunnerSeccompProfile", () => {
     }
   });
 });
+
+describe("runnerSecurityContext", () => {
+  it("sets nothing when off", () => {
+    expect(runnerSecurityContext("off", 1000, "node-24")).toBeUndefined();
+  });
+
+  it("drops all capabilities and blocks escalation at baseline", () => {
+    expect(runnerSecurityContext("baseline", 1000, "node-24")).toEqual({
+      allowPrivilegeEscalation: false,
+      capabilities: { drop: ["ALL"] },
+    });
+  });
+
+  it("pins the configured uid when restricted", () => {
+    expect(runnerSecurityContext("restricted", 1000, "node-24")).toEqual({
+      allowPrivilegeEscalation: false,
+      capabilities: { drop: ["ALL"] },
+      runAsNonRoot: true,
+      runAsUser: 1000,
+    });
+  });
+
+  it("pins bun's own uid, which differs from node's", () => {
+    expect(runnerSecurityContext("restricted", 1000, "bun")?.runAsUser).toBe(1001);
+  });
+
+  it("falls back to the configured uid when the runtime is unknown", () => {
+    for (const runtime of [undefined, null, "", "node", "node-22", "node-26"]) {
+      expect(runnerSecurityContext("restricted", 1000, runtime)?.runAsUser).toBe(1000);
+    }
+  });
+});

apps/supervisor/src/workloadManager/kubernetes.ts

@@ -17,6 +17,7 @@ import { getRunnerId } from "../util.js";
 import {
   nodetypeNodeSelector,
   runPodTolerations,
+  runnerSecurityContext,
   withRunnerSeccompProfile,
   withNodeSelector,
 } from "./kubernetesPodSpec.js";
@@ -175,6 +176,11 @@ export class KubernetesWorkloadManager implements WorkloadManager {
                   },
                 ],
                 resources: this.#getResourcesForMachine(opts.machine),
+                securityContext: runnerSecurityContext(
+                  env.KUBERNETES_RUNNER_SECURITY_CONTEXT,
+                  env.KUBERNETES_RUNNER_RUN_AS_USER,
+                  opts.runtime
+                ),
                 env: [
                   {
                     name: "TRIGGER_DEQUEUED_AT_MS",

apps/supervisor/src/workloadManager/kubernetesPodSpec.ts

@@ -96,3 +96,33 @@ export function withRunnerSeccompProfile(
     },
   };
 }
+
+const BUN_RUN_AS_USER = 1001;
+
+/**
+ * runnerSecurityContext maps a configured level onto the run container's security
+ * context. "baseline" drops the capability bounding set and blocks setuid
+ * escalation; "restricted" additionally pins the container to a non-root uid.
+ *
+ * The uid is set explicitly rather than read from the image: the kubelet cannot
+ * verify `runAsNonRoot` against an image that declares a named user, and fails
+ * the container instead. Bun images carry their user at a different uid to
+ * node's, so the runtime selects which uid is pinned.
+ */
+export function runnerSecurityContext(
+  level: "off" | "baseline" | "restricted",
+  runAsUser: number,
+  runtime: string | null | undefined
+): k8s.V1SecurityContext | undefined {
+  if (level === "off") {
+    return undefined;
+  }
+
+  return {
+    allowPrivilegeEscalation: false,
+    capabilities: { drop: ["ALL"] },
+    ...(level === "restricted"
+      ? { runAsNonRoot: true, runAsUser: runtime === "bun" ? BUN_RUN_AS_USER : runAsUser }
+      : {}),
+  };
+}

apps/webapp/app/components/billing/OrgBanner.tsx

@@ -157,14 +157,14 @@ function NoLimitConfiguredBanner() {
             to={v3BillingLimitsPath(organization)}
           >
             <span className="mx-auto grow self-center truncate text-text-bright system:text-white">
-              Configure billing limit
+              Billing limit settings
             </span>
           </LinkButton>
         ) : undefined
       }
     >
       {canManageBillingLimits
-        ? "Protect your organization from unexpected usage spikes."
+        ? "Add a billing limit to your account to prevent overspending"
         : "Billing limits are not configured for this organization. Contact an organization administrator to configure them."}
     </AnimatedOrgBannerBar>
   );