I'm searching for recommendations on open-source tools or platforms that provide full observability for Kubernetes. I need options that cover metrics, logs, traces, alerting, dashboards, and more. I'm eager to hear what tools you use and what you'd recommend. Thanks!
5 Answers
In my larger setup with thousands of nodes, we depend on the Prometheus-Thanos combo for metrics, and for logging and tracing, we use Filebeat with Kafka, Logstash, and Elasticsearch. Grafana and Kibana help us with visualization based on our needs.
If you're exploring options, check out this list of vendors sorted by open-source offerings. Some options include Apache SkyWalking, Elastic, Grafana LGTM, and more. Not every tool covers all features, but you can find a variety of choices!
I have issues with Elastic for metrics; it ends up being much bulkier than Prometheus. Plus, without enterprise options, it doesn't integrate well with Grafana.
I've been using the LGTM stack and it's pretty solid! However, I'm leaning towards incorporating VictoriaMetrics and possibly VictoriaLogs soon. OpenTelemetry has been great for ingestion too—it feels very stable and user-friendly.
I've had a great experience with VictoriaMetrics as well, but I've noticed some limitations in Grafana's drill-down options when using VictoriaLogs and traces.
I'm on board with VictoriaMetrics and VictoriaLogs too, and I'm quite satisfied with their performance.
For a robust observability setup, I use Grafana with VictoriaMetrics, VictoriaLogs, and Tempo. Alerts go through Grafana with PagerDuty for on-call notifications, plus I use various exporters to gather metrics.
If you're on a budget, the LGTM stack is great! Using Prometheus for node and cluster utilization along with Calico for network observability really helps, along with Grafana for dashboards. If you're considering spending, Calico enterprise could be worth it.
But LGTM offers similar functionalities and I’ve found Mimir to outperform Thanos in performance.