[APM] Should this promise be retried?

[afharo]

The promises below might fail in case there's a glitch in ES connection and never retry.

kibana/x-pack/plugins/observability_solution/apm/server/lib/apm_telemetry/index.ts

Lines 120 to 149 in 593d391

	await taskManagerStart.ensureScheduled({
	id: APM_TELEMETRY_TASK_NAME,
	taskType: APM_TELEMETRY_TASK_NAME,
	schedule: {
	interval: '720m',
	},
	scope: ['apm'],
	params: {},
	state: {},
	});
	try {
	const currentData = (
	await savedObjectsClient.get(
	APM_TELEMETRY_SAVED_OBJECT_TYPE,
	APM_TELEMETRY_SAVED_OBJECT_ID
	)
	).attributes as { kibanaVersion?: string };
	if (currentData.kibanaVersion !== kibanaVersion) {
	logger.debug(
	`Stored telemetry is out of date. Task will run immediately. Stored: ${currentData.kibanaVersion}, expected: ${kibanaVersion}`
	);
	await taskManagerStart.runSoon(APM_TELEMETRY_TASK_NAME);
	}
	} catch (err) {
	if (!SavedObjectsErrorHelpers.isNotFoundError(err)) {
	logger.warn('Failed to fetch saved telemetry data.');
	}
	}

kibana/x-pack/plugins/observability_solution/apm/server/plugin.ts

Lines 71 to 79 in 593d391

	createApmTelemetry({
	core,
	getApmIndices: plugins.apmDataAccess.getApmIndices,
	usageCollector: plugins.usageCollection,
	taskManager: plugins.taskManager,
	logger: this.logger,
	kibanaVersion: this.initContext.env.packageInfo.version,
	isProd: this.initContext.env.mode.prod,
	}).catch(() => {});

kibana/x-pack/plugins/observability_solution/apm/server/plugin.ts

Lines 130 to 141 in 593d391

	apmIndicesPromise
	.then((apmIndices) => {
	plugins.home?.tutorials.registerTutorial(
	tutorialProvider({
	apmConfig: currentConfig,
	apmIndices,
	cloud: plugins.cloud,
	isFleetPluginEnabled: !isEmpty(resourcePlugins.fleet),
	})
	);
	})
	.catch(() => {});

Is that OK? Should we apply any retry logic?

If we want to retry, p-retry could be a useful utility that's already present in the project.

[elasticmachine]

Pinging @elastic/apm-ui (Team:APM)

[dgieselaar]

@afharo It's okay to not retry. We want to keep this as unintrusive as possible, to prevent any unnecessary load on the cluster.

[afharo]

@dgieselaar, thanks for confirming! FWIW, I created this issue for awareness. Closing the issue is perfectly valid.

Just to confirm about this case

kibana/x-pack/plugins/observability_solution/apm/server/plugin.ts

Lines 130 to 141 in 593d391

	apmIndicesPromise
	.then((apmIndices) => {
	plugins.home?.tutorials.registerTutorial(
	tutorialProvider({
	apmConfig: currentConfig,
	apmIndices,
	cloud: plugins.cloud,
	isFleetPluginEnabled: !isEmpty(resourcePlugins.fleet),
	})
	);
	})
	.catch(() => {});

If the APM promise fails, the plugin won't register the tutorial

Wrt telemetry-related promises, if they fail, we won't have telemetry. But I understand we might be ok with that to avoid overloading the cluster.

[dgieselaar]

@afharo ah sorry missed the last one! I think a retry makes sense there, but I will leave it up to @elastic/obs-ux-infra_services-team to figure out what is best there. Thanks for bringing it up in any case!