mshade | June 8, 2023, 2:12 p.m.
I've long been looking for a simple(r) tool for observability in small clusters, especially something that doesn't require 10 different components to simply keep an eye on things. The ELK stack is large and complex, and Prometheus + Grafana takes a lot of horsepower and storage space. Third party tools like DataDog and NewRelic are limited with their free tier and grow expensive quickly for longer retention time periods or more custom metrics.
An interesting newcomer to the space has announced its release, OpenObserve -- promising decreased storage costs and a much simpler platform that combines log, metric, and trace visibility into one app.
The simplest installation is a single deployment, containerized or native, and uses local storage with the sled database. The HA deployment model uses etcd and S3 or other compatible blob-storage.
I've kicked the tires on OpenObserve for log parsing and searching in my home lab, and though the UI is a little unpolished, the simplicity is definitely appreciated! I'm shipping logs with fluent-bit and configuration couldn't have been simpler.
The quickstart install guide provides several methods for getting started, but I opted to create a helm chart for single-node deployment for added flexibility over the provided kubernetes manifest. It creates a single statefulset replica with configurable ingress, storage, and can set a randomly generated initial root password to avoid hard-coding secrets into the environment as the provided example manifest does.