fix(sessions): skip durable fsync for session store

steipete · steipete · commit 897bac5b8cd3 · 2026-05-07T12:31:18.000+01:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -10,6 +10,7 @@ Docs: https://docs.openclaw.ai
 - Discord/voice: keep TTS playback running when another user starts speaking, ignore new capture during playback to avoid feedback loops, and downgrade expected receive-stream aborts to verbose diagnostics.
 - Telegram: treat successful same-chat `message` tool outbound sends during an inbound telegram turn as delivered when deciding whether to emit the rewritten silent reply fallback (#78685). Thanks @neeravmakwana.
 - Gateway/tasks: reconcile stale CLI run-context tasks whose live run context disappeared even when a child session row remains, and apply the default bounded reload deferral timeout to channel hot reloads so stale task records cannot block Discord/Slack/Telegram reloads forever.
+- Gateway/sessions: keep session-store index writes atomic while skipping durable fsync inside the writer lock, reducing cron and channel-turn starvation on slow filesystems. Fixes #73655. Thanks @mmartoccia.
 - Discord/voice: make `openclaw channels capabilities --channel discord --target channel:<id>` and `channels status --probe` audit voice-channel permissions, including auto-join targets, so missing Connect/Speak/Read Message History permissions show up before `/vc join`.
 - Channels CLI: make `openclaw channels list` channel-only — drop the `Auth providers (OAuth + API keys)` block (use `openclaw models auth list`), drop the per-provider usage/quota fetch and the `--no-usage` flag (use `openclaw status` or `openclaw models list`), add `--all` to surface bundled-unconfigured, catalog-not-installed, and catalog-installed-but-unconfigured channels, and render explicit `installed` / `configured` / `enabled` tags per row plus an `origin` + `installed` field in JSON. Fixes WeCom-class catalog channels disappearing from `--all` when installed on disk but not yet configured. (#78456) Thanks @sliverp.
 - CLI/cron: add computed `status` field to `cron list --json` and `cron show <id> --json` output, mirroring the human-readable status column (disabled/running/ok/error/skipped/idle) so external tooling can determine job state without re-deriving it from raw state fields. (#78701) Thanks @aweiker.
diff --git a/src/config/sessions/sessions.test.ts b/src/config/sessions/sessions.test.ts
@@ -337,6 +337,31 @@ describe("session store writer queue", () => {
     writeSpy.mockRestore();
   });
 
+  it("keeps session store writes atomic while skipping durable fsync inside the writer lock", async () => {
+    const key = "agent:main:no-fsync";
+    const { storePath } = await makeTmpStore({
+      [key]: { sessionId: "s-no-fsync", updatedAt: Date.now(), counter: 0 },
+    });
+
+    const writeSpy = vi.spyOn(jsonFiles, "writeTextAtomic");
+    await updateSessionStore(
+      storePath,
+      async (store) => {
+        const entry = store[key] as Record<string, unknown>;
+        entry.counter = 1;
+      },
+      { skipMaintenance: true },
+    );
+
+    expect(writeSpy).toHaveBeenCalledTimes(1);
+    expect(writeSpy).toHaveBeenCalledWith(
+      storePath,
+      expect.any(String),
+      expect.objectContaining({ durable: false, mode: 0o600 }),
+    );
+    writeSpy.mockRestore();
+  });
+
   it("multiple consecutive errors do not permanently poison the queue", async () => {
     const key = "agent:main:multi-err";
     const { storePath } = await makeTmpStore({
diff --git a/src/config/sessions/store.ts b/src/config/sessions/store.ts
@@ -499,7 +499,7 @@ async function writeSessionStoreAtomic(params: {
   store: Record<string, SessionEntry>;
   serialized: string;
 }): Promise<void> {
-  await writeTextAtomic(params.storePath, params.serialized, { mode: 0o600 });
+  await writeTextAtomic(params.storePath, params.serialized, { durable: false, mode: 0o600 });
   updateSessionStoreWriteCaches({
     storePath: params.storePath,
     store: params.store,
