fix(runner): accept valid large journals during session continuation

Co-Authored-By: Paperclip <noreply@paperclip.ing>

packages/paperclip-runner/src/control-plane/durable-prp-control-plane.ts

@@ -57,7 +57,9 @@ const maxCommands = 500;
 // polling turn regains the event loop. Match the transport's explicit deferred
 // event bound so a valid burst is not compacted before it can be observed.
 const maxCommittedEventWindow = 4_096;
-const maxStateBytes = 192 * 1024 * 1024;
+// The controller admission, cleanup, and recovery readers inspect this same
+// journal. Keep their bound aligned with the durable store as history grows.
+export const DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES = 192 * 1024 * 1024;
 const authChallengeTtlMs = 5_000;
 const stableIdPattern = /^[A-Za-z0-9][A-Za-z0-9._:-]{0,239}$/;
 const runnerDigestPattern = /^sha256:[0-9a-f]{64}$/;
@@ -1113,7 +1115,7 @@ function readPrivateFile(path: string): string | null {
   try {
     const metadata = fstatSync(descriptor);
     verifyPrivateRegularFile(metadata, path);
-    if (metadata.size > maxStateBytes) {
+    if (metadata.size > DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES) {
       throw new Error(`Private state file exceeds its size bound: ${path}`);
     }
     return readFileSync(descriptor, "utf8");

packages/paperclip-runner/src/index.ts

@@ -27,6 +27,7 @@ export {
 export * from "./native-session-runtime.js";
 export {
   DurablePrpControlPlane,
+  DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
   inspectWarmRunTransition,
   type DurablePrpControlPlaneOptions,
   type PrpWireConnection,

server/src/services/native-runtime/native-session-executor.test.ts

@@ -38,6 +38,8 @@ import { DatabaseSync } from "node:sqlite";
 import { nativeSha256 } from "./canonical.js";
 import * as noLaunchProofModule from "./native-maintenance-no-launch.js";
 import {
+  DurablePrpControlPlane,
+  DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
   NativeSessionCleanupQuarantinedError,
   NativeProviderTerminalFailure,
   NativeSessionProtocolIntegrityError,
@@ -8704,6 +8706,108 @@ describe("runnerd provider runtime wiring", () => {
     },
   );
 
+  it.each([
+    { commandCount: 4, fault: null },
+    { commandCount: 24, fault: null },
+    { commandCount: 4, fault: "oversized" },
+    { commandCount: 4, fault: "wrong identity" },
+  ])(
+    "admits a valid large control-plane journal while preserving bounds and identity ($commandCount commands, $fault)",
+    async ({ commandCount, fault }) => {
+      const stateBase = await mkdtemp(join(tmpdir(), "paperclip-large-prp-journal-"));
+      const previousStateDirectory = process.env.PAPERCLIP_RUNNER_STATE_DIR;
+      process.env.PAPERCLIP_RUNNER_STATE_DIR = stateBase;
+      const prior = {
+        ...execution,
+        binding: { ...execution.binding, runId: "large-journal-prior" },
+      } as NativeExecutionInputV1;
+      const current = {
+        ...prior,
+        binding: { ...prior.binding, runId: "large-journal-next" },
+      } as NativeExecutionInputV1;
+      const identity = {
+        runId: prior.binding.runId,
+        normalizedSessionId: prior.session.normalizedSessionId!,
+        runnerInstanceId: "large-journal-runner",
+        environmentLeaseId: "large-journal-lease",
+        turnId: "large-journal-turn",
+        itemId: "large-journal-item",
+      };
+      const remoteTarget = {
+        kind: "remote", transport: "sandbox", providerKey: "daytona",
+        leaseId: identity.environmentLeaseId,
+        remoteCwd: "/home/daytona/paperclip-workspace",
+        runner: { execute: vi.fn() },
+      } as never;
+      const priorRunDb = {
+        select: () => ({ from: () => ({ where: () => ({ limit: () => Promise.resolve([{
+          status: "succeeded", runnerProfileJson: { nativeExecutionInput: prior },
+        }]) }) }) }),
+      } as unknown as Db;
+      try {
+        state.createBackend.mockClear();
+        state.createTransport.mockClear();
+        await createRunnerdBackend({
+          db: leaseDb(prior), execution: prior,
+          runnerInstanceId: identity.runnerInstanceId, runnerExecutionTarget: remoteTarget,
+        });
+        state.createBackend.mock.calls[0]![1].codexTransportFactory!();
+        const root = state.createTransport.mock.calls[0]![0].stateDirectory!;
+        const options = {
+          stateDirectory: join(root, "control-plane"), identity,
+          expectedRunnerVersion: "0.3.0", expectedRunnerDigest: `sha256:${"a".repeat(64)}`,
+        };
+        // The real writer accepts and reloads these individually bounded commands.
+        // Accumulated history must not invalidate an otherwise identical session.
+        const core = new DurablePrpControlPlane(options);
+        for (let index = 0; index < commandCount; index += 1) {
+          core.queueCommand("turn.start", { text: "x".repeat(768 * 1024) }, `large-command-${index}`);
+        }
+        for (const command of core.store.state.commands) command.status = "completed";
+        core.store.save();
+        expect(new DurablePrpControlPlane(options).store.state.commands).toHaveLength(commandCount);
+        const controlPath = join(root, "control-plane", "control-plane-state.json");
+        const controlBytes = await readFile(controlPath);
+        expect(controlBytes.byteLength).toBeGreaterThan((commandCount === 4 ? 2 : 16) * 1024 * 1024);
+        await mkdir(join(root, "runner"), { recursive: true });
+        await writeFile(join(root, "runner", "runner-state.json"), JSON.stringify(durableRunnerState(identity, "suspended")));
+        if (fault === "oversized") {
+          await truncate(controlPath, DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES + 1);
+        } else if (fault === "wrong identity") {
+          await writeFile(controlPath, JSON.stringify({
+            ...core.store.state,
+            identity: { ...identity, normalizedSessionId: "unrelated-session" },
+          }));
+        }
+        state.createBackend.mockClear();
+        state.createTransport.mockClear();
+        const continuation = createRunnerdBackend({
+          db: priorRunDb, execution: current, runnerInstanceId: "new-heartbeat-runner",
+          runnerExecutionTarget: remoteTarget,
+        });
+        if (fault !== null) {
+          await expect(continuation).rejects.toThrow("runner_state_identity_mismatch");
+          expect(state.createBackend).not.toHaveBeenCalled();
+          expect(state.createTransport).not.toHaveBeenCalled();
+          return;
+        }
+        await expect(continuation).resolves.toBeDefined();
+        state.createBackend.mock.calls[0]![1].codexTransportFactory!();
+        expect(state.createTransport.mock.calls[0]![0].prpIdentity).toMatchObject({
+          runId: current.binding.runId,
+          runnerInstanceId: identity.runnerInstanceId,
+          environmentLeaseId: identity.environmentLeaseId,
+        });
+        expect((await readFile(controlPath)).equals(controlBytes)).toBe(true);
+        await expect(access(join(stateBase, "quarantine"))).rejects.toThrow();
+      } finally {
+        if (previousStateDirectory === undefined) delete process.env.PAPERCLIP_RUNNER_STATE_DIR;
+        else process.env.PAPERCLIP_RUNNER_STATE_DIR = previousStateDirectory;
+        await rm(stateBase, { recursive: true, force: true });
+      }
+    },
+  );
+
   it("quarantines legacy prior-run state only after the database proves a terminal owner in the same full scope", async () => {
     const stateBase = await mkdtemp(
       join(tmpdir(), "paperclip-legacy-terminal-unsuspended-state-"),
@@ -8901,6 +9005,7 @@ describe("runnerd provider runtime wiring", () => {
     "quarantined",
     "empty retry shell",
     "unsuspended current",
+    "large control journal",
     "live runner",
     "live group",
     "missing process identity",
@@ -9089,13 +9194,17 @@ describe("runnerd provider runtime wiring", () => {
             join(root, "control-plane", "control-plane-state.json"),
             JSON.stringify({
               ...durableControlPlaneState(identity),
-              commands: [
-                {
+              commands: Array.from(
+                { length: scenario === "large control journal" ? 24 : 1 },
+                () => ({
                   type: "turn.start",
+                  payload: scenario === "large control journal"
+                    ? { text: "x".repeat(768 * 1024) }
+                    : {},
                   status:
                     scenario === "pending command" ? "pending" : "completed",
-                },
-              ],
+                }),
+              ),
               committedEvents: [
                 { eventType: "run.terminal", envelope: identity },
               ],
@@ -9138,6 +9247,7 @@ describe("runnerd provider runtime wiring", () => {
           "quarantined",
           "empty retry shell",
           "unsuspended current",
+          "large control journal",
           "active goal",
         ].includes(scenario);
         if (shouldRecover) {

server/src/services/native-runtime/native-session-executor.ts

@@ -63,6 +63,7 @@ import type {
   PrpStructuredRunResult,
 } from "../../vendor/paperclip-runner/index.js";
 import {
+  DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
   NativeProviderTerminalFailure,
   NativeSessionCleanupQuarantinedError,
   NativeSessionProtocolIntegrityError,
@@ -276,7 +277,6 @@ export async function detachNativeSessionsForRestart(
 const MAX_REMOTE_CHECKPOINT_ARCHIVE_BYTES = 64 * 1024 * 1024;
 const MAX_REMOTE_CHECKPOINT_EXPANDED_BYTES = 64 * 1024 * 1024;
 const MAX_REMOTE_CHECKPOINT_ENTRIES = 20_000;
-const NATIVE_DURABLE_IDENTITY_MAX_BYTES = 2 * 1024 * 1024;
 const NATIVE_RUNNER_STATE_MAX_BYTES = 16 * 1024 * 1024;
 const NATIVE_WARM_CHECKPOINT_MAX_BYTES = 8 * 1024 * 1024;
 const CODEX_HOME_NON_PERSISTENT_ENTRIES = [
@@ -1775,6 +1775,12 @@ const CLEANUP_CANONICAL_FILES = [
 ] as const;
 const CLEANUP_ACTIVATION_FILE = "cleanup-activation.json";
 
+function nativeStateFileMaxBytes(relativePath: string): number {
+  return relativePath === "control-plane/control-plane-state.json"
+    ? DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES
+    : NATIVE_RUNNER_STATE_MAX_BYTES;
+}
+
 function cleanupStateSnapshot(root: string, providerFile = "codex-provider-state.json") {
   if (
     ![root, resolve(root, "runner"), resolve(root, "control-plane")].every(
@@ -1787,7 +1793,7 @@ function cleanupStateSnapshot(root: string, providerFile = "codex-provider-state
   const bytes = files.map((file) =>
     readBoundedNativeFile(
       resolve(root, file),
-      NATIVE_RUNNER_STATE_MAX_BYTES,
+      nativeStateFileMaxBytes(file),
       "native_cleanup_maintenance_unproven",
     ),
   );
@@ -4060,7 +4066,7 @@ function hasRetainedWarmTransitionEvidence(root: string): boolean {
     [
       "control-plane",
       "control-plane-state.json",
-      NATIVE_DURABLE_IDENTITY_MAX_BYTES,
+      DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
     ],
     ["runner", "runner-state.json", NATIVE_RUNNER_STATE_MAX_BYTES],
   ] as const) {
@@ -4117,7 +4123,7 @@ function readWarmTransitionSnapshot(root: string) {
   }
   const core = readBoundedNativeFile(
     resolve(root, "control-plane", "control-plane-state.json"),
-    NATIVE_DURABLE_IDENTITY_MAX_BYTES,
+    DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
     "native_runner_warm_transition_recovery_unproven",
   );
   const runner = readBoundedNativeFile(
@@ -4718,7 +4724,7 @@ async function recoverQuiescentRunnerdState(input: {
           throw new Error("unsafe_recovery_state");
         return readBoundedNativeFile(
           resolve(root, directory, name),
-          NATIVE_RUNNER_STATE_MAX_BYTES,
+          nativeStateFileMaxBytes(`${directory}/${name}`),
           "recovery_state_too_large",
         ).toString("utf8");
       };
@@ -4895,7 +4901,7 @@ async function recoverQuiescentRunnerdState(input: {
     if (
       readBoundedNativeFile(
         resolve(candidate.root, relativePath!),
-        NATIVE_RUNNER_STATE_MAX_BYTES,
+        nativeStateFileMaxBytes(relativePath!),
         "recovery_state_too_large",
       ).toString("utf8") !== expected
     ) {
@@ -4958,7 +4964,7 @@ export function runnerdStateProvesIncompleteBootstrap(root: string): boolean {
       JSON.parse(
         readBoundedNativeFile(
           statePath,
-          NATIVE_DURABLE_IDENTITY_MAX_BYTES,
+          DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
           "runner_durable_identity_too_large",
         ).toString("utf8"),
       ),
@@ -5011,7 +5017,7 @@ function readRunnerdDurableIdentity(
       JSON.parse(
         readBoundedNativeFile(
           statePath,
-          NATIVE_DURABLE_IDENTITY_MAX_BYTES,
+          DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES,
           "runner_durable_identity_too_large",
         ).toString("utf8"),
       ),

server/src/vendor/paperclip-runner/index.ts

@@ -68,6 +68,8 @@ const sourceUrl = new URL(
 const runner = (await import(sourceUrl.href)) as RunnerModule;
 
 export const DurablePrpControlPlane = runner.DurablePrpControlPlane;
+export const DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES =
+  runner.DURABLE_PRP_CONTROL_PLANE_MAX_STATE_BYTES;
 export const runnerCodexDynamicToolsFit = runner.runnerCodexDynamicToolsFit;
 export const inspectWarmRunTransition = runner.inspectWarmRunTransition;
 export const readRunnerdArtifactBinding = runner.readRunnerdArtifactBinding;