Transcript

Alright, you just deployed your back end application. Traffic is coming in, but how do you actually know if the system is working? Not just whether it's up, but how it's performing, how it's handling load, and if it's behaving the way you expect. That's where metrics come in. Metrics and measurement your system emits over time. That tells you what's happening under the hood. Think of them as your systems vital signs. Just like a doctor checks your heart rate, blood pressure and temperature to understand your health, you check the metrics to understand the health of your systems. The first layer is the infrastructure metrics, which is the. Foundation. This is where you monitor the physical machine that your application is running on. Here you might have metrics to track CPU usage, memory consumption, disk IO, and network throughput. These are your baselines. If anyone of these is off by a lot, there could be potential problems and it usually shows up here before customers start complaining. The second layer is application metrics. This is 1 level up. You're no longer looking at the physical machine, you're looking at your application itself. The metrics you might track here might include how many requests are coming in, how many of those requests are succeeding. Versus failing of the request that are failing, you might want to know what percentage are four hundreds, which are client errors, or five hundreds which are internal server errors. You might also want to track your error rate over time. You could also monitor your latency for your upstream callers and for applications that you call downstream as well. These metrics tell you whether your application is serving traffic correctly. The third layer is custom business or behavioral metrics. This is where things get specific to what your system actually does. Say your service is supposed to process payments, You might want to track things like how many payments. Are being initiated versus completed, how long the checkout flow is taking end to end, or how many retries are happening on a particular operation. These are metrics that know out-of-the-box tool will define for you because they are unique to your applications behavior. Now metrics alone aren't enough. You typically pair them with alerting. So when something crosses the threshold you've defined, you get notified to investigate before it becomes a major incident. As always For The Woman calls for more programming videos like this and check out womancalls.com for more.