Investigate whether renderTime metric can be replaced with ELB checks

[cemms1]

In our regular dashboard session for WebX, we often check the renderTime percentiles Kibana graph which uses the time taken to render a page based on logging info. This graph takes a long time to load and we aren't sure if we get much value from it.

PR which introduced logging rendering times: #2721

Permalink:

dotcom-rendering/dotcom-rendering/src/server/prod-server.ts

Lines 32 to 40 in 0528f48

	// Middleware to track route performance using 'response-time' lib
	// Usage: app.post('/Article', logRenderTime, renderArticle);
	const logRenderTime = responseTime(
	(_1: Request, _2: Response, renderTime: number) => {
	logger.info('Page render time', {
	renderTime,
	});
	},
	);

We might be duplicating efforts here as the load balancer should give us the request time for DCR. Using the data from the ELB should be much faster than checking logging information.

The result of this ticket should confirm whether or not we can remove the render time logging in favour of ELB data. If so, we should remove the chart in Kibana and remove the logging in DCR, unless we need to use it for any other reason.

[jamesgorrie]

This might make it hard to slice the data by endpoint, which has proven useful in the past.

Saying that, this may become less of an issue if we split the stacks with their own ELBs:
#8351

I'm not 100% sure what meta data get's attached to AWS logs? But pretty certain request information isn't attached.