CCIP-5216 Tracking latencies and simplification in the metrics code

[mateusz-sekara]

Simplifying tracking metrics for the processors by:

• reducing number of interfaces and moving stats to the struct
• reducing number of method in the reporter
• tracking metrics is done in the decorator (see TrackedProcessor), so plugin is less polluted with the observability stuff
• any processor can add tracking of their observations and outcomes as long it's wrapped with TrackedProcessor

Additionally, tracking errors and latencies for the processor methods

[0xnogo]

Are the rootsLabel, messagesLabel, rmnSignatureLabel under the processorLabel? Meaning all the 3 labels will be dimensions of merkleroot metric?

[mateusz-sekara]

I don't understand the question (btw these labels were there before, I'm only moving them from one place to another)

[dimkouv]

Let's make sure we track durations for individual processor states.
E.g.

labels = {
   "plugin": "commit",
   "processor": "merkleRoots",
   "phase": "observation",
   "state": "buildingReport",
   "duration": x
}

labels = {
   "plugin": "exec",
   "phase": "observation",
   "state": "filtering"
   "duration": x
}

It is also useful to track the oracleID in this metrics.
Sometimes maybe a specific oracle has issues that contribute to high P99,P90 because the RPC they are using is slow.

[mateusz-sekara]

It is also useful to track the oracleID in this metrics.
Sometimes maybe a specific oracle has issues that contribute to high P99,P90 because the RPC they are using is slow.

These are tracked automatically

[github-actions]

Metric	detailed-metrics	main
Coverage	74.7%	75.3%