fix: prevent stale timeout from killing actively running background tasks
The stale detection was checking lastUpdate timestamps BEFORE consulting session.status(), causing tasks to be unfairly killed after 3 minutes even when the session was actively running (e.g., during long tool executions or extended thinking). Changes: - Reorder pollRunningTasks to fetch session.status() before stale check - Skip stale-kill entirely when session status is 'running' - Port no-lastUpdate handling from task-poller.ts into manager.ts (previously manager silently skipped tasks without lastUpdate) - Add sessionStatuses parameter to checkAndInterruptStaleTasks - Add 7 new test cases covering session-status-aware stale detection
This commit is contained in:
YeonGyu-Kim
@@ -2289,10 +2289,221 @@ describe("BackgroundManager.checkAndInterruptStaleTasks", () => {
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
await manager["checkAndInterruptStaleTasks"]()
|
||||
await manager["checkAndInterruptStaleTasks"]()
|
||||
|
||||
expect(task.status).toBe("cancelled")
|
||||
})
|
||||
|
||||
test("should NOT interrupt task when session is running, even with stale lastUpdate", async () => {
|
||||
//#given
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { staleTimeoutMs: 180_000 })
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-running-session",
|
||||
sessionID: "session-running",
|
||||
parentSessionID: "parent-rs",
|
||||
parentMessageID: "msg-rs",
|
||||
description: "Task with running session",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 300_000),
|
||||
progress: {
|
||||
toolCalls: 2,
|
||||
lastUpdate: new Date(Date.now() - 300_000),
|
||||
},
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — session is actively running
|
||||
await manager["checkAndInterruptStaleTasks"]({ "session-running": { type: "running" } })
|
||||
|
||||
//#then — task survives because session is running
|
||||
expect(task.status).toBe("running")
|
||||
})
|
||||
|
||||
test("should interrupt task when session is idle and lastUpdate exceeds stale timeout", async () => {
|
||||
//#given
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { staleTimeoutMs: 180_000 })
|
||||
stubNotifyParentSession(manager)
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-idle-session",
|
||||
sessionID: "session-idle",
|
||||
parentSessionID: "parent-is",
|
||||
parentMessageID: "msg-is",
|
||||
description: "Task with idle session",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 300_000),
|
||||
progress: {
|
||||
toolCalls: 2,
|
||||
lastUpdate: new Date(Date.now() - 300_000),
|
||||
},
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — session is idle
|
||||
await manager["checkAndInterruptStaleTasks"]({ "session-idle": { type: "idle" } })
|
||||
|
||||
//#then — killed because session is idle with stale lastUpdate
|
||||
expect(task.status).toBe("cancelled")
|
||||
expect(task.error).toContain("Stale timeout")
|
||||
})
|
||||
|
||||
test("should NOT interrupt running session even with very old lastUpdate (no safety net)", async () => {
|
||||
//#given
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { staleTimeoutMs: 180_000 })
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-long-running",
|
||||
sessionID: "session-long",
|
||||
parentSessionID: "parent-lr",
|
||||
parentMessageID: "msg-lr",
|
||||
description: "Long running task",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 900_000),
|
||||
progress: {
|
||||
toolCalls: 5,
|
||||
lastUpdate: new Date(Date.now() - 900_000),
|
||||
},
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — session is running, lastUpdate 15min old
|
||||
await manager["checkAndInterruptStaleTasks"]({ "session-long": { type: "running" } })
|
||||
|
||||
//#then — running sessions are NEVER stale-killed
|
||||
expect(task.status).toBe("running")
|
||||
})
|
||||
|
||||
test("should NOT interrupt running session with no progress (undefined lastUpdate)", async () => {
|
||||
//#given — no progress at all, but session is running
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { messageStalenessTimeoutMs: 600_000 })
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-running-no-progress",
|
||||
sessionID: "session-rnp",
|
||||
parentSessionID: "parent-rnp",
|
||||
parentMessageID: "msg-rnp",
|
||||
description: "Running no progress",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 15 * 60 * 1000),
|
||||
progress: undefined,
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — session is running despite no progress
|
||||
await manager["checkAndInterruptStaleTasks"]({ "session-rnp": { type: "running" } })
|
||||
|
||||
//#then — running sessions are NEVER killed
|
||||
expect(task.status).toBe("running")
|
||||
})
|
||||
|
||||
test("should interrupt task with no lastUpdate after messageStalenessTimeout", async () => {
|
||||
//#given
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { messageStalenessTimeoutMs: 600_000 })
|
||||
stubNotifyParentSession(manager)
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-no-update",
|
||||
sessionID: "session-no-update",
|
||||
parentSessionID: "parent-nu",
|
||||
parentMessageID: "msg-nu",
|
||||
description: "No update task",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 15 * 60 * 1000),
|
||||
progress: undefined,
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — no progress update for 15 minutes
|
||||
await manager["checkAndInterruptStaleTasks"]({})
|
||||
|
||||
//#then — killed after messageStalenessTimeout
|
||||
expect(task.status).toBe("cancelled")
|
||||
expect(task.error).toContain("no activity")
|
||||
})
|
||||
|
||||
test("should NOT interrupt task with no lastUpdate within messageStalenessTimeout", async () => {
|
||||
//#given
|
||||
const client = {
|
||||
session: {
|
||||
prompt: async () => ({}),
|
||||
promptAsync: async () => ({}),
|
||||
abort: async () => ({}),
|
||||
},
|
||||
}
|
||||
const manager = new BackgroundManager({ client, directory: tmpdir() } as unknown as PluginInput, { messageStalenessTimeoutMs: 600_000 })
|
||||
|
||||
const task: BackgroundTask = {
|
||||
id: "task-fresh-no-update",
|
||||
sessionID: "session-fresh",
|
||||
parentSessionID: "parent-fn",
|
||||
parentMessageID: "msg-fn",
|
||||
description: "Fresh no-update task",
|
||||
prompt: "Test",
|
||||
agent: "test-agent",
|
||||
status: "running",
|
||||
startedAt: new Date(Date.now() - 5 * 60 * 1000),
|
||||
progress: undefined,
|
||||
}
|
||||
|
||||
getTaskMap(manager).set(task.id, task)
|
||||
|
||||
//#when — only 5 min since start, within 10min timeout
|
||||
await manager["checkAndInterruptStaleTasks"]({})
|
||||
|
||||
//#then — task survives
|
||||
expect(task.status).toBe("running")
|
||||
})
|
||||
})
|
||||
|
||||
describe("BackgroundManager.shutdown session abort", () => {
|
||||
|
||||
Reference in New Issue
Block a user