November 14, 2025
Hold my logs: dashboard drama
I can't recommend Grafana anymore
From dashboard darling to drama: users slam Grafana churn
TLDR: An engineer says Grafana’s constant product changes and broken dashboards make it hard to recommend. The crowd is split: some say it still works fine if you don’t chase updates, others complain about nonstop rebuilds, and many admit there’s no clear open-source replacement—cue calls for a community fork.
One engineer’s breakup post with Grafana hit a nerve: after years of love—easy dashboards, neat logs, cloud freebies—the vibes turned sour as features got deprecated and dashboards broke. The author says Grafana kept swapping parts: OnCall gone, Agent replaced by Alloy, old charts abandoned, and a switch in the dashboard tech that shattered existing setups. Translation for non-tech folks: the furniture keeps getting moved, and everyone trips over the coffee table.
The comments? A spicy split. The top mood is upgrade fatigue: people groan that every few years tools go poof and they’re left rebuilding. One user called it a “replace-a-thon,” likening it to tech Jenga where “every quarter” something big breaks. Yet defenders pipe up: one team runs Grafana a version behind and says “everything still works,” praising its Swiss‑army‑knife approach to data sources. Then there’s the despairing middle: if not Grafana, what else? Some asked for a fork—open-source slang for “let’s split off and make our own”—while another joked the site’s font color is so faint they can’t even read the complaints. Peak tech drama.
Underneath the memes, a real fear: Grafana’s push to be a full platform (hello, Alloy) feels like whiplash. Fans want stability; skeptics want out; everyone wants fewer surprise rebuilds.
Key Points
- •Initial observability stack used Loki, Prometheus, and Grafana via docker-compose, with Promtail for non-Docker logs and minimal overhead.
- •High-cardinality labels in Loki can exhaust filesystem inodes due to Cortex bin-packing, a lesson learned by the author.
- •Kubernetes migration introduced storage challenges for Prometheus and a need for 13-month retention, leading to evaluation of Thanos vs. Mimir; Mimir was chosen.
- •Grafana’s product landscape changed rapidly: Grafana Agent moved to Flow Mode and was later deprecated; Grafana OnCall and some Helm charts were also deprecated or unmaintained.
- •Grafana Alloy was announced as the unified replacement for Grafana Agent, supporting logs, metrics, traces, and OpenTelemetry; it had early issues but improved, and uses an HCL-like config language.