refactor: clear restart sentinels from sqlite

This commit is contained in:
Peter Steinberger
2026-05-08 17:06:26 +01:00
parent 840b420a3b
commit 9204b488d2
10 changed files with 28 additions and 94 deletions

View File

@@ -380,8 +380,8 @@ The remaining cleanup is mostly consolidation and deletion:
- Gateway singleton locks now use shared SQLite KV instead of temp-dir lock
files. Done.
- Gateway restart sentinel state now uses shared SQLite KV instead of
`restart-sentinel.json`; the old path resolver remains only for legacy
cleanup/status compatibility.
`restart-sentinel.json`; runtime code clears the SQLite row directly and no
longer carries file cleanup plumbing.
- Gateway restart intent and supervisor handoff state now use shared SQLite KV
instead of `gateway-restart-intent.json` and
`gateway-supervisor-restart-handoff.json` sidecars.

View File

@@ -6,10 +6,10 @@ import { createGatewayTool } from "./gateway-tool.js";
type ScheduleGatewayRestartArgs = Parameters<typeof scheduleGatewaySigusr1Restart>[0];
const {
clearRestartSentinelMock,
extractDeliveryInfoMock,
formatDoctorNonInteractiveHintMock,
isRestartEnabledMock,
removeRestartSentinelFileMock,
scheduleGatewaySigusr1RestartMock,
writeRestartSentinelMock,
} = vi.hoisted(() => ({
@@ -23,8 +23,8 @@ const {
threadId: "thread-42",
})),
formatDoctorNonInteractiveHintMock: vi.fn(() => "Run: openclaw doctor --non-interactive"),
writeRestartSentinelMock: vi.fn(async (_payload: RestartSentinelPayload) => "/tmp/restart"),
removeRestartSentinelFileMock: vi.fn(async (_path: string | null | undefined) => undefined),
writeRestartSentinelMock: vi.fn(async (_payload: RestartSentinelPayload) => undefined),
clearRestartSentinelMock: vi.fn(async () => undefined),
scheduleGatewaySigusr1RestartMock: vi.fn((_opts?: ScheduleGatewayRestartArgs) => ({
scheduled: true,
delayMs: 250,
@@ -46,7 +46,7 @@ vi.mock("../../infra/restart-sentinel.js", async () => {
return {
...actual,
formatDoctorNonInteractiveHint: formatDoctorNonInteractiveHintMock,
removeRestartSentinelFile: removeRestartSentinelFileMock,
clearRestartSentinel: clearRestartSentinelMock,
writeRestartSentinel: writeRestartSentinelMock,
};
});
@@ -98,8 +98,8 @@ describe("gateway tool restart continuation", () => {
formatDoctorNonInteractiveHintMock.mockReset();
formatDoctorNonInteractiveHintMock.mockReturnValue("Run: openclaw doctor --non-interactive");
writeRestartSentinelMock.mockReset();
writeRestartSentinelMock.mockResolvedValue("/tmp/restart");
removeRestartSentinelFileMock.mockClear();
writeRestartSentinelMock.mockResolvedValue(undefined);
clearRestartSentinelMock.mockClear();
scheduleGatewaySigusr1RestartMock.mockReset();
scheduleGatewaySigusr1RestartMock.mockReturnValue({ scheduled: true, delayMs: 250 });
});
@@ -224,6 +224,6 @@ describe("gateway tool restart continuation", () => {
await scheduledArgs?.emitHooks?.beforeEmit?.();
await scheduledArgs?.emitHooks?.afterEmitRejected?.();
expect(removeRestartSentinelFileMock).toHaveBeenCalledWith("/tmp/restart");
expect(clearRestartSentinelMock).toHaveBeenCalledOnce();
});
});

View File

@@ -7,8 +7,8 @@ import { extractDeliveryInfo } from "../../config/sessions.js";
import type { OpenClawConfig } from "../../config/types.openclaw.js";
import {
buildRestartSuccessContinuation,
clearRestartSentinel,
formatDoctorNonInteractiveHint,
removeRestartSentinelFile,
type RestartSentinelPayload,
writeRestartSentinel,
} from "../../infra/restart-sentinel.js";
@@ -414,16 +414,15 @@ export function createGatewayTool(opts?: {
log.info(
`gateway tool: restart requested (delayMs=${delayMs ?? "default"}, reason=${reason ?? "none"})`,
);
let sentinelPath: string | null = null;
const scheduled = scheduleGatewaySigusr1Restart({
delayMs,
reason,
emitHooks: {
beforeEmit: async () => {
sentinelPath = await writeRestartSentinel(payload);
await writeRestartSentinel(payload);
},
afterEmitRejected: async () => {
await removeRestartSentinelFile(sentinelPath);
await clearRestartSentinel();
},
},
});

View File

@@ -14,8 +14,8 @@ import { getSessionBindingService } from "../../infra/outbound/session-binding-s
import type { SessionBindingRecord } from "../../infra/outbound/session-binding-service.js";
import {
buildRestartSuccessContinuation,
clearRestartSentinel,
formatDoctorNonInteractiveHint,
removeRestartSentinelFile,
type RestartSentinelPayload,
writeRestartSentinel,
} from "../../infra/restart-sentinel.js";
@@ -695,16 +695,15 @@ export const handleRestartCommand: CommandHandler = async (params, allowTextComm
const hasSigusr1Listener = process.listenerCount("SIGUSR1") > 0;
const sentinelPayload = buildRestartCommandSentinel(params);
if (hasSigusr1Listener) {
let sentinelPath: string | null = null;
scheduleGatewaySigusr1Restart({
reason: "/restart",
emitHooks: sentinelPayload
? {
beforeEmit: async () => {
sentinelPath = await writeRestartSentinel(sentinelPayload);
await writeRestartSentinel(sentinelPayload);
},
afterEmitRejected: async () => {
await removeRestartSentinelFile(sentinelPath);
await clearRestartSentinel();
},
}
: undefined,
@@ -716,10 +715,9 @@ export const handleRestartCommand: CommandHandler = async (params, allowTextComm
},
};
}
let sentinelPath: string | null = null;
try {
if (sentinelPayload) {
sentinelPath = await writeRestartSentinel(sentinelPayload);
await writeRestartSentinel(sentinelPayload);
}
} catch (err) {
logVerbose(`failed to write /restart sentinel: ${String(err)}`);
@@ -732,7 +730,7 @@ export const handleRestartCommand: CommandHandler = async (params, allowTextComm
}
const restartMethod = triggerOpenClawRestart();
if (!restartMethod.ok) {
await removeRestartSentinelFile(sentinelPath);
await clearRestartSentinel();
const detail = restartMethod.detail ? ` Details: ${restartMethod.detail}` : "";
return {
shouldContinue: false,

View File

@@ -17,9 +17,7 @@ const scheduleGatewaySigusr1RestartMock = vi.fn(() => ({
coalesced: false,
}));
const restartSentinelMocks = vi.hoisted(() => ({
writeRestartSentinel: vi.fn(async (_payload: RestartSentinelPayload) => {
return "/tmp/restart-sentinel.json";
}),
writeRestartSentinel: vi.fn(async (_payload: RestartSentinelPayload) => undefined),
}));
vi.mock("../../config/config.js", async () => {

View File

@@ -29,8 +29,7 @@ const mocks = vi.hoisted(() => {
},
},
})),
removeRestartSentinelFile: vi.fn(async () => undefined),
resolveRestartSentinelPath: vi.fn(() => "/tmp/restart-sentinel.json"),
clearRestartSentinel: vi.fn(async () => undefined),
formatRestartSentinelMessage: vi.fn(() => "restart message"),
summarizeRestartSentinel: vi.fn(() => "restart summary"),
resolveMainSessionKeyFromConfig: vi.fn(() => "agent:main:main"),
@@ -169,8 +168,7 @@ vi.mock("../agents/agent-scope.js", async () => {
vi.mock("../infra/restart-sentinel.js", () => ({
readRestartSentinel: mocks.readRestartSentinel,
removeRestartSentinelFile: mocks.removeRestartSentinelFile,
resolveRestartSentinelPath: mocks.resolveRestartSentinelPath,
clearRestartSentinel: mocks.clearRestartSentinel,
formatRestartSentinelMessage: mocks.formatRestartSentinelMessage,
summarizeRestartSentinel: mocks.summarizeRestartSentinel,
}));
@@ -344,7 +342,7 @@ describe("scheduleRestartSentinelWake", () => {
mocks.loadPendingSessionDelivery.mockClear();
mocks.drainPendingSessionDeliveries.mockClear();
mocks.recoverPendingSessionDeliveries.mockClear();
mocks.removeRestartSentinelFile.mockClear();
mocks.clearRestartSentinel.mockClear();
mocks.injectTimestamp.mockClear();
mocks.timestampOptsFromConfig.mockClear();
mocks.recordInboundSessionAndDispatchReply.mockReset();
@@ -1131,7 +1129,7 @@ describe("scheduleRestartSentinelWake", () => {
await scheduleRestartSentinelWake({ deps: {} as never });
expect(mocks.removeRestartSentinelFile).not.toHaveBeenCalled();
expect(mocks.clearRestartSentinel).not.toHaveBeenCalled();
expect(mocks.drainPendingSessionDeliveries).not.toHaveBeenCalled();
expect(mocks.logWarn).toHaveBeenCalledWith(
"startup task failed",

View File

@@ -15,13 +15,12 @@ import { ackDelivery, enqueueDelivery, failDelivery } from "../infra/outbound/de
import { buildOutboundSessionContext } from "../infra/outbound/session-context.js";
import { resolveOutboundTarget } from "../infra/outbound/targets.js";
import {
clearRestartSentinel,
finalizeUpdateRestartSentinelRunningVersion,
formatRestartSentinelMessage,
readRestartSentinel,
removeRestartSentinelFile,
type RestartSentinelContinuation,
type RestartSentinelPayload,
resolveRestartSentinelPath,
summarizeRestartSentinel,
} from "../infra/restart-sentinel.js";
import {
@@ -476,7 +475,6 @@ async function loadRestartSentinelStartupTask(params: {
if (!sentinel) {
return null;
}
const sentinelPath = resolveRestartSentinelPath();
const payload = sentinel.payload;
const sessionKey = payload.sessionKey?.trim();
const message = formatRestartSentinelMessage(payload);
@@ -498,7 +496,7 @@ async function loadRestartSentinelStartupTask(params: {
continuationKind: payload.continuation.kind,
});
}
await removeRestartSentinelFile(sentinelPath);
await clearRestartSentinel();
return { status: "ran" as const };
}
@@ -591,7 +589,7 @@ async function loadRestartSentinelStartupTask(params: {
);
}
await removeRestartSentinelFile(sentinelPath);
await clearRestartSentinel();
const routedAgentTurnContinuation =
payload.continuation?.kind === "agentTurn" && continuationRoute !== undefined;
if (!routedAgentTurnContinuation) {

View File

@@ -50,14 +50,6 @@ describe("gateway restart intent", () => {
expect(fs.existsSync(intentPath(env))).toBe(false);
});
it("rejects oversized intent files before parsing", () => {
const env = createIntentEnv();
fs.writeFileSync(intentPath(env), "x".repeat(2048), { encoding: "utf8", mode: 0o600 });
expect(consumeGatewayRestartIntentSync(env)).toBe(false);
expect(fs.existsSync(intentPath(env))).toBe(true);
});
it("stores intents in SQLite instead of a legacy JSON file", () => {
const env = createIntentEnv();
@@ -84,21 +76,4 @@ describe("gateway restart intent", () => {
});
expect(fs.existsSync(intentPath(env))).toBe(false);
});
it("does not touch an existing legacy intent-path symlink when writing", () => {
const env = createIntentEnv();
const targetPath = path.join(env.OPENCLAW_STATE_DIR ?? "", "attacker-target.txt");
fs.writeFileSync(targetPath, "keep", "utf8");
try {
fs.symlinkSync(targetPath, intentPath(env));
} catch {
return;
}
expect(writeGatewayRestartIntentSync({ env, targetPid: process.pid })).toBe(true);
expect(fs.readFileSync(targetPath, "utf8")).toBe("keep");
expect(fs.lstatSync(intentPath(env)).isSymbolicLink()).toBe(true);
expect(consumeGatewayRestartIntentSync(env)).toBe(true);
});
});

View File

@@ -1,4 +1,3 @@
import fs from "node:fs/promises";
import { describe, expect, it } from "vitest";
import { closeOpenClawStateDatabaseForTest } from "../state/openclaw-state-db.js";
import {
@@ -17,7 +16,6 @@ import {
formatRestartSentinelMessage,
markUpdateRestartSentinelFailure,
readRestartSentinel,
resolveRestartSentinelPath,
summarizeRestartSentinel,
trimLogTail,
writeRestartSentinel,
@@ -53,9 +51,7 @@ describe("restart sentinel", () => {
},
stats: { mode: "git" },
};
const filePath = await writeRestartSentinel(payload);
expect(filePath).toBe(resolveRestartSentinelPath());
await expect(fs.stat(filePath)).rejects.toThrow();
await writeRestartSentinel(payload);
const read = await readRestartSentinel();
expect(read?.payload.kind).toBe("update");
@@ -70,17 +66,6 @@ describe("restart sentinel", () => {
});
});
it("ignores legacy sentinel files at runtime", async () => {
await withRestartSentinelStateDir(async () => {
const filePath = resolveRestartSentinelPath();
await fs.writeFile(filePath, "not-json", "utf-8");
const read = await readRestartSentinel();
expect(read).toBeNull();
await expect(fs.readFile(filePath, "utf-8")).resolves.toBe("not-json");
});
});
it("drops structurally invalid SQLite sentinel payloads", async () => {
await withRestartSentinelStateDir(async () => {
writeOpenClawStateKvJson<OpenClawStateJsonValue>(

View File

@@ -1,7 +1,4 @@
import fs from "node:fs/promises";
import path from "node:path";
import { formatCliCommand } from "../cli/command-format.js";
import { resolveStateDir } from "../config/paths.js";
import {
deleteOpenClawStateKvJson,
readOpenClawStateKvJson,
@@ -71,7 +68,6 @@ export type RestartSentinel = {
export const DEFAULT_RESTART_SUCCESS_CONTINUATION_MESSAGE =
"The gateway restart completed successfully. Tell the user OpenClaw restarted successfully and continue any pending work.";
const SENTINEL_FILENAME = "restart-sentinel.json";
const RESTART_SENTINEL_KV_SCOPE = "gateway.restart-sentinel";
const RESTART_SENTINEL_KV_KEY = "current";
@@ -81,10 +77,6 @@ export function formatDoctorNonInteractiveHint(
return `Run: ${formatCliCommand("openclaw doctor --non-interactive", env)}`;
}
export function resolveRestartSentinelPath(env: NodeJS.ProcessEnv = process.env): string {
return path.join(resolveStateDir(env), SENTINEL_FILENAME);
}
export async function writeRestartSentinel(
payload: RestartSentinelPayload,
env: NodeJS.ProcessEnv = process.env,
@@ -96,7 +88,6 @@ export async function writeRestartSentinel(
data as unknown as OpenClawStateJsonValue,
{ env },
);
return resolveRestartSentinelPath(env);
}
function isPlainRecord(value: unknown): value is Record<string, unknown> {
@@ -163,15 +154,8 @@ export async function markUpdateRestartSentinelFailure(
}, env);
}
export async function removeRestartSentinelFile(
filePath: string | null | undefined,
env: NodeJS.ProcessEnv = process.env,
) {
export async function clearRestartSentinel(env: NodeJS.ProcessEnv = process.env) {
deleteOpenClawStateKvJson(RESTART_SENTINEL_KV_SCOPE, RESTART_SENTINEL_KV_KEY, { env });
if (!filePath) {
return;
}
await fs.unlink(filePath).catch(() => {});
}
export function buildRestartSuccessContinuation(params: {
@@ -211,12 +195,11 @@ export async function hasRestartSentinel(env: NodeJS.ProcessEnv = process.env):
export async function consumeRestartSentinel(
env: NodeJS.ProcessEnv = process.env,
): Promise<RestartSentinel | null> {
const filePath = resolveRestartSentinelPath(env);
const parsed = await readRestartSentinel(env);
if (!parsed) {
return null;
}
await removeRestartSentinelFile(filePath, env);
await clearRestartSentinel(env);
return parsed;
}