Feat/performance tracking

[gsteenkamp89]

closes ACX-3198

Goal:
Provide a standardised way of logging performance metrics.
The API is similar to using Node's native perf_hooks, but slightly easier to use.

The Profiler will ensure { datadog: true } is appended to all logs so we can ingest this by searching for this key.

Since the Profiler class holds some state, each instance should be created within a function scope so it can be garbage collected.

async function functionToMeasure( logger: Logger ) {
   const profiler = new Profiler({
      logger,
      at: "Relayer::functionToMeasure"
   })

  // drop first mark
  profiler.mark("A");

  // do stuff
  await step1();

  profiler.mark("B", { depositId: 123 });

  // do stuff
  await step2();

  profiler.mark("C", { relayerAddress: "0xabcabc1231231" });

  // Measure durations between marks
  profiler.measure("task1", {
    from: "A",
    to: "B",
    message: "Time from task A to task B",
  });

  profiler.measure("task2", {
    from: "B",
    to: "C",
    message: "Time from task B to task C",
  });
  profiler.measure("Total", {
    from: "A",
    to: "C",
    message: "Time from task A to task C",
  });
}

or wrap an async function with measureAsync

async function calculateFoo(...args: any): Promise<Foo> {
   return Foo
}

const foo = await profiler.measureAsync(calculateFoo(...args), "calculateFoo", {
   message: "Time taken to calculate Foo",
   ...args // log additional data
});

[pxrl]

Great to see this - I left a few comments; interested to see what you think.

[linear]

ACX-3198 bots: Pattern/strategy for tracking performance of arbitrary components

[nicholaspai]

where is the logic for this?

The Profiler will ensure { datadog: true } is appended to all logs

This looks like a great approach, I like the class' interface a lot

[gsteenkamp89]

where is the logic for this?

The Profiler will ensure { datadog: true } is appended to all logs

This looks like a great approach, I like the class' interface a lot

the default metadata { datadog: true } is being set here

[pxrl]

@gsteenkamp89 LGTM. Pending that one question on the logger, are you OK to re-propose Profiler in the SDK? Then we can keep this PR open and use it to update/align all the uses of the profiler in the relayer.

[james-a-morris]

Is there any overlap here? across-protocol/indexer#95

[gsteenkamp89]

Is there any overlap here? across-protocol/indexer#95

Oh jeez.. Yes there's definitely overlap. Although looking at David's PR, it seems the goal there as slightly different maybe? This profiler is just logging production performance data to datadog.
The benchmarker is generating reports on demand or on a schedule which is maybe more useful in a CI job as a part of testing our code in the development cycle.
🤔