fix(observability): bound orphaned-run finalization POST during shutdown (SDK-7518)

.changeset/pr-180.md

@@ -0,0 +1,5 @@
+---
+"@wdio/browserstack-service": patch
+---
+
+- Fixed builds occasionally showing as "running" for hours after the run finished, caused by an unbounded network request during shutdown.

packages/browserstack-service/src/constants.ts

@@ -256,6 +256,12 @@ export const STOP_BUILD_ATTEMPT_TIMEOUT_MS = 10000
 export const STOP_BUILD_TOTAL_BUDGET_MS = 30000
 export const STOP_BUILD_BACKOFF_BASE_MS = 1000
 
+// SDK-7518: finalizeOrphanedRuns posts synthetic TestRunFinished events during shutdown,
+// immediately before the build-stop call. Bound that POST so a hung connection cannot stall
+// shutdown before the build is stopped (which leaves the build "running" until the server-side
+// inactivity timeout). Sibling to STOP_BUILD_ATTEMPT_TIMEOUT_MS on the same shutdown path.
+export const ORPHAN_FINALIZE_POST_TIMEOUT_MS = 10000
+
 // API Endpoint constants
 export const UPDATED_CLI_ENDPOINT = 'sdk/v1/update_cli'
 

packages/browserstack-service/src/testOps/openRunsJournal.ts

@@ -3,7 +3,7 @@ import fs from 'node:fs'
 
 import type { TestData, UploadType } from '../types.js'
 import { batchAndPostEvents } from '../util.js'
-import { DATA_BATCH_ENDPOINT } from '../constants.js'
+import { DATA_BATCH_ENDPOINT, ORPHAN_FINALIZE_POST_TIMEOUT_MS } from '../constants.js'
 import { BStackLogger } from '../bstackLogger.js'
 
 /**
@@ -94,7 +94,7 @@ export async function finalizeOrphanedRuns(): Promise<number> {
             }
             return { event_type: 'TestRunFinished', test_run: finishedRun }
         })
-        await batchAndPostEvents(DATA_BATCH_ENDPOINT, 'ORPHANED_TEST_RUN_FINALIZATION', events)
+        await batchAndPostEvents(DATA_BATCH_ENDPOINT, 'ORPHANED_TEST_RUN_FINALIZATION', events, ORPHAN_FINALIZE_POST_TIMEOUT_MS)
         BStackLogger.info(`Finalized ${events.length} orphaned test/hook run(s) left behind by an interrupted run`)
         return events.length
     } catch (e) {

packages/browserstack-service/src/util.ts

@@ -1327,7 +1327,7 @@ export function shouldAddServiceVersion(config: Options.Testrunner, testObservab
     return true
 }
 
-export async function batchAndPostEvents (eventUrl: string, kind: string, data: UploadType[]) {
+export async function batchAndPostEvents (eventUrl: string, kind: string, data: UploadType[], timeoutMs?: number) {
     if (!process.env[TESTOPS_BUILD_COMPLETED_ENV]) {
         throw new Error('Build not completed yet')
     }
@@ -1337,6 +1337,12 @@ export async function batchAndPostEvents (eventUrl: string, kind: string, data:
         throw new Error('Missing authentication Token')
     }
 
+    // SDK-7518: an optional per-call timeout. When set (by shutdown-path callers such as
+    // finalizeOrphanedRuns), bound the fetch with an AbortController so a hung connection cannot
+    // stall shutdown before the build-stop call. Left unset by normal in-run callers, whose
+    // behaviour is unchanged.
+    const controller = timeoutMs ? new AbortController() : undefined
+    const timeoutId = controller ? setTimeout(() => controller.abort(), timeoutMs) : undefined
     try {
         const url = `${APIUtils.DATA_ENDPOINT}/${eventUrl}`
         const response = await fetch(url, {
@@ -1345,7 +1351,8 @@ export async function batchAndPostEvents (eventUrl: string, kind: string, data:
                 ...DEFAULT_REQUEST_CONFIG.headers,
                 'Authorization': `Bearer ${jwtToken}`
             },
-            body: JSON.stringify(data)
+            body: JSON.stringify(data),
+            signal: controller?.signal
         })
         // read as text first: error responses (401/5xx) and empty bodies are not JSON, and a blind
         // response.json() surfaced them as a misleading "Unexpected end of JSON input"
@@ -1357,6 +1364,8 @@ export async function batchAndPostEvents (eventUrl: string, kind: string, data:
     } catch (error) {
         BStackLogger.debug(`[${kind}] EXCEPTION IN ${kind} REQUEST TO TEST REPORTING AND ANALYTICS : ${error}`)
         throw new Error('Exception in request ' + error)
+    } finally {
+        if (timeoutId) { clearTimeout(timeoutId) }
     }
 }