From 2d72a3811981eea4cad0dcaeade1aad2ce8ebca4 Mon Sep 17 00:00:00 2001
From: Gerard Soldevila <gerard.soldevila@elastic.co>
Date: Thu, 14 Dec 2023 17:31:58 +0100
Subject: [PATCH] Inject initial "not reported" to prevent blocking
 `/api/status` (#173371)

## Summary

Addresses https://github.com/elastic/kibana-team/issues/697

The endpoint is currently waiting for all statuses to be reported.
[Recent changes](https://github.com/elastic/kibana/pull/172268) are
causing the `pluginsStatus$` Observable to take longer to emit first
value.

This PR injects a "status not reported" to prevent the different status
Observables from blocking calls to the `/api/status` enpdoint.
---
 .../src/routes/status.ts                      | 29 ++++++++++++++-----
 .../http/platform/status.ts                   |  6 ++--
 2 files changed, 24 insertions(+), 11 deletions(-)

diff --git a/packages/core/status/core-status-server-internal/src/routes/status.ts b/packages/core/status/core-status-server-internal/src/routes/status.ts
index e06d667b4c78b..59c7aa23c51d4 100644
--- a/packages/core/status/core-status-server-internal/src/routes/status.ts
+++ b/packages/core/status/core-status-server-internal/src/routes/status.ts
@@ -6,7 +6,7 @@
  * Side Public License, v 1.
  */
 
-import { Observable, combineLatest, ReplaySubject, firstValueFrom } from 'rxjs';
+import { type Observable, combineLatest, ReplaySubject, firstValueFrom, startWith } from 'rxjs';
 import { schema } from '@kbn/config-schema';
 import type { PackageInfo } from '@kbn/config';
 import type { PluginName } from '@kbn/core-base-common';
@@ -15,12 +15,12 @@ import type { MetricsServiceSetup } from '@kbn/core-metrics-server';
 import type { CoreIncrementUsageCounter } from '@kbn/core-usage-data-server';
 import type { StatusResponse } from '@kbn/core-status-common-internal';
 import {
-  ServiceStatus,
-  ServiceStatusLevel,
-  CoreStatus,
+  type ServiceStatus,
+  type ServiceStatusLevel,
+  type CoreStatus,
   ServiceStatusLevels,
 } from '@kbn/core-status-common';
-import { calculateLegacyStatus, LegacyStatusInfo } from '../legacy_status';
+import { calculateLegacyStatus, type LegacyStatusInfo } from '../legacy_status';
 
 const SNAPSHOT_POSTFIX = /-SNAPSHOT$/;
 
@@ -61,6 +61,11 @@ export interface RedactedStatusHttpBody {
   };
 }
 
+const SERVICE_UNAVAILABLE_NOT_REPORTED: ServiceStatus = {
+  level: ServiceStatusLevels.unavailable,
+  summary: 'Status not yet reported',
+};
+
 export const registerStatusRoute = ({
   router,
   config,
@@ -73,9 +78,17 @@ export const registerStatusRoute = ({
   const combinedStatus$ = new ReplaySubject<
     [ServiceStatus<unknown>, ServiceStatus, CoreStatus, Record<string, ServiceStatus<unknown>>]
   >(1);
-  combineLatest([status.overall$, status.coreOverall$, status.core$, status.plugins$]).subscribe(
-    combinedStatus$
-  );
+  combineLatest([
+    status.overall$.pipe(startWith(SERVICE_UNAVAILABLE_NOT_REPORTED)),
+    status.coreOverall$.pipe(startWith(SERVICE_UNAVAILABLE_NOT_REPORTED)),
+    status.core$.pipe(
+      startWith<CoreStatus>({
+        elasticsearch: SERVICE_UNAVAILABLE_NOT_REPORTED,
+        savedObjects: SERVICE_UNAVAILABLE_NOT_REPORTED,
+      })
+    ),
+    status.plugins$.pipe(startWith({})),
+  ]).subscribe(combinedStatus$);
 
   router.get(
     {
diff --git a/test/server_integration/http/platform/status.ts b/test/server_integration/http/platform/status.ts
index 8ce193c670849..df472ec36d818 100644
--- a/test/server_integration/http/platform/status.ts
+++ b/test/server_integration/http/platform/status.ts
@@ -35,13 +35,13 @@ export default function ({ getService }: FtrProviderContext) {
     // This test must come first because the timeout only applies to the initial emission
     it("returns a timeout for status check that doesn't emit after 30s", async () => {
       let aStatus = await getStatus('statusPluginA');
-      expect(aStatus.level).to.eql('unavailable');
+      expect(aStatus === undefined || aStatus.level === 'unavailable').to.eql(true);
 
       // Status will remain in unavailable until the custom status check times out
       // Keep polling until that condition ends, up to a timeout
       await retry.waitForWithTimeout(`Status check to timeout`, 40_000, async () => {
         aStatus = await getStatus('statusPluginA');
-        return aStatus.summary === 'Status check timed out after 30s';
+        return aStatus?.summary === 'Status check timed out after 30s';
       });
 
       expect(aStatus.level).to.eql('unavailable');
@@ -53,7 +53,7 @@ export default function ({ getService }: FtrProviderContext) {
       await retry.waitForWithTimeout(
         `statusPluginA status to update`,
         5_000,
-        async () => (await getStatus('statusPluginA')).level === 'degraded'
+        async () => (await getStatus('statusPluginA'))?.level === 'degraded'
       );
       await statusPropagation();
       expect((await getStatus('statusPluginA')).level).to.eql('degraded');