From 0e2775f678918bf45b6ef2f8ef5bf36942c92a78 Mon Sep 17 00:00:00 2001
From: Lola <omolola.akinleye@elastic.co>
Date: Wed, 13 Sep 2023 10:35:50 -0400
Subject: [PATCH] [Cloud Security][Telemetry] add error handling incase an
 individual collector fails (#165918)

## Summary

Summarize your PR. If it involves visual changes include a screenshot or
gif.
The kibana `cloud security` usage collector payload fails when one
feature collector fails. I applied error handling and enabled the
feature collector to fail gracefully with a default value. We are now
using `Promise.allSettled` vs `Promse.all` to avoid all collectors
failing in case one collector fails.
---
 .../installation_stats_collector.ts           |  2 +-
 .../lib/telemetry/collectors/register.ts      | 42 +++++++++++--------
 .../server/lib/telemetry/collectors/types.ts  |  8 ++++
 3 files changed, 34 insertions(+), 18 deletions(-)

diff --git a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/installation_stats_collector.ts b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/installation_stats_collector.ts
index 09b197245cd5a4..038ed35abac97d 100644
--- a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/installation_stats_collector.ts
+++ b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/installation_stats_collector.ts
@@ -25,7 +25,7 @@ const getEnabledInputStreamVars = (packagePolicy: PackagePolicy) => {
 const getAccountTypeField = (
   packagePolicy: PackagePolicy
 ): CloudSecurityInstallationStats['account_type'] => {
-  if (packagePolicy.vars?.posture.value !== 'cspm') return;
+  if (packagePolicy.vars?.posture?.value !== 'cspm') return;
 
   const inputStreamVars = getEnabledInputStreamVars(packagePolicy);
   const cloudProvider = packagePolicy.vars?.deployment?.value;
diff --git a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/register.ts b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/register.ts
index c9495c03eccdb9..f41497cd80bb4e 100644
--- a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/register.ts
+++ b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/register.ts
@@ -11,7 +11,7 @@ import { CspServerPluginStart, CspServerPluginStartDeps } from '../../../types';
 import { getIndicesStats } from './indices_stats_collector';
 import { getResourcesStats } from './resources_stats_collector';
 import { cspmUsageSchema } from './schema';
-import { CspmUsage } from './types';
+import { CspmUsage, type CloudSecurityUsageCollectorType } from './types';
 import { getAccountsStats } from './accounts_stats_collector';
 import { getRulesStats } from './rules_stats_collector';
 import { getInstallationStats } from './installation_stats_collector';
@@ -35,6 +35,22 @@ export function registerCspmUsageCollector(
       return true;
     },
     fetch: async (collectorFetchContext: CollectorFetchContext) => {
+      const awaitPromiseSafe = async <T>(
+        taskName: CloudSecurityUsageCollectorType,
+        promise: Promise<T>
+      ) => {
+        try {
+          const val = await promise;
+          return val;
+        } catch (error) {
+          logger.error(`${taskName} task failed: ${error.message}`);
+          logger.error(error.stack);
+          return error;
+        }
+      };
+
+      const esClient = collectorFetchContext.esClient;
+      const soClient = collectorFetchContext.soClient;
       const [
         indicesStats,
         accountsStats,
@@ -43,24 +59,16 @@ export function registerCspmUsageCollector(
         installationStats,
         alertsStats,
       ] = await Promise.all([
-        getIndicesStats(
-          collectorFetchContext.esClient,
-          collectorFetchContext.soClient,
-          coreServices,
-          logger
+        awaitPromiseSafe('Indices', getIndicesStats(esClient, soClient, coreServices, logger)),
+        awaitPromiseSafe('Accounts', getAccountsStats(esClient, logger)),
+        awaitPromiseSafe('Resources', getResourcesStats(esClient, logger)),
+        awaitPromiseSafe('Rules', getRulesStats(esClient, logger)),
+        awaitPromiseSafe(
+          'Installation',
+          getInstallationStats(esClient, soClient, coreServices, logger)
         ),
-        getAccountsStats(collectorFetchContext.esClient, logger),
-        getResourcesStats(collectorFetchContext.esClient, logger),
-        getRulesStats(collectorFetchContext.esClient, logger),
-        getInstallationStats(
-          collectorFetchContext.esClient,
-          collectorFetchContext.soClient,
-          coreServices,
-          logger
-        ),
-        getAlertsStats(collectorFetchContext.esClient, logger),
+        awaitPromiseSafe('Alerts', getAlertsStats(esClient, logger)),
       ]);
-
       return {
         indices: indicesStats,
         accounts_stats: accountsStats,
diff --git a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/types.ts b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/types.ts
index 0c04de498509a5..f1162e24d3168d 100644
--- a/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/types.ts
+++ b/x-pack/plugins/cloud_security_posture/server/lib/telemetry/collectors/types.ts
@@ -7,6 +7,14 @@
 
 import { CspStatusCode } from '../../../../common/types';
 
+export type CloudSecurityUsageCollectorType =
+  | 'Indices'
+  | 'Accounts'
+  | 'Resources'
+  | 'Rules'
+  | 'Installation'
+  | 'Alerts';
+
 export interface CspmUsage {
   indices: CspmIndicesStats;
   resources_stats: CspmResourcesStats[];