Topic

Observability & Monitoring

You can't fix what you can't see — but you also can't fix what's drowning you in alerts. Prometheus, Grafana, Loki, Tempo, and Alertmanager set up so they actually help instead of paging you about a normal Tuesday. SLO basics, alert fatigue, log volume that won't melt your disks, and the dashboards worth keeping versus the ones that just look busy.

68 articles in this topic.

Featured posts

Langfuse vs Phoenix: Self-Host LLM Tracing

Langfuse v4 runs six containers and idles near 1.1 GiB of RAM. Phoenix v21 runs two and idles near 0.5 GiB. Here is what each traces and what each costs you.

8 Nov, 2026
12 min read
Recording Rules: Faster Prometheus Panels

Prometheus recording rules precompute the expensive queries behind your Grafana panels. Learn naming, chaining, testing with promtool, and when to skip them.

28 Oct, 2026
14 min read
NVMe SMART: Reading Wear Before It Dies

How NVMe wear leveling works and which SMART health log fields show it: Percentage Used, Available Spare, media errors, TBW math, smartctl vs nvme-cli.

24 Oct, 2026
13 min read
Uptime Kuma vs Gatus vs Cachet

Uptime Kuma, Gatus, and Cachet all make a status page, but only two of them check anything. Here is which one to run in a home lab, with Compose files for each.

14 Oct, 2026
13 min read
What Actually Fills Your Loki Storage

Your Loki volume hit tens of GB and you don't know why. Find the noisy streams with logcli and LogQL, then trim them in Alloy and set per-stream retention.

13 Oct, 2026
13 min read
Grafana Dashboard Sprawl: 50 Panels to 10

Your Grafana dashboard is a wall of 50 panels nobody reads. Use USE and RED to cut it to ten, move the rest to drill-downs, and keep the result in git.

6 Oct, 2026
13 min read