Understanding cloud observability
Software delivery culture is evolving. It is moving from monitoring to observability. This is particularly important for microservice and/or cloud-native applications. It is essential to monitor their behavior and make sure they are working properly at all times.
"The massive migration to the cloud, which was already taking place at a steady pace before the pandemic, will reach a peak in 2021," Forrester pointed out at the end of October. Another confirmation from this analyst firm: next year, 60% of (large) enterprises worldwide will be using containers on public cloud platforms.
More than ever, the cloud is becoming the norm. But its adoption presents new operational challenges. Today's microservices environments, deployed on container orchestration platforms such as Kubernetes, are based on scalable, ephemeral containers.
To understand the interactions between the components of your system, it is therefore essential to follow the life cycle of a request in distributed environments. In this context, observability must be treated as a prime concern in operations planning.
It offers many advantages, such as improved productivity and performance (allocating resources efficiently according to user needs), but also optimal detection to resolve problems (undesirable events) before they affect your business.
The concept of microservice observability is similar to FCAPS (Fault, configuration, accounting, performance, security) in the traditional world of telecoms. When applied to the cloud, it can be divided into three categories:
Application performance management
Metrics are useful for understanding the behavior of a system. Whether purely technical or business-oriented, they help to make code ever more maintainable and easy to use.
In the Kubernetes ecosystem, metrics are available at many levels. There are several solutions, but Prometheus is the benchmark for monitoring cloud-native resources. Its automated installation and configuration through the operator makes it the benchmark tool for monitoring a Kubernetes cluster.
Logs
They are considered one of the main pillars of observability, as they enable us to see what actually happens to applications and microservices in the native cloud environment. Cloud providers typically use the so-called EFK stack - Elasticsearch, Fluentd and Kibana - to record logs and generate alerts.
Tracing
This section provides a better understanding of the complete lifecycle of a request. Tracing is both practical and essential for identifying problems, such as increased latency or excessive use of resources.
But tracking the exchanges between each microservice instance generates an almost unmanageable amount of data. As a result, monitoring tools only collect a very small sample for analysis and visualization. But even this small sample is useful, as it allows us to see how microservices work.
On the surface, observability and monitoring have similarities. To avoid misunderstandings, think of monitoring as a mechanic servicing your car when a maintenance light comes on. On a server, it's about ensuring that users can access your applications and services without worry.
Like vehicle maintenance, notifications are sent to teams when something isn't working properly or a malfunction is likely to occur.
But before sending a notification, it's essential to create a baseline and understand the current state of the environment. And that's where observability comes in. Without observing the architecture, it's impossible to know what's wrong and what's wrong.
