At last count my homelab Grafana has thirty-one dashboards. I have, generously, about a dozen services worth monitoring. The maths there is not flattering, and I noticed it because when something actually went wrong last week, I could not find the one panel that would have told me what, amongst the thirty others I had lovingly built and then forgotten.
This is the trap with self-hosted monitoring. Adding a dashboard is fun. It scratches the same itch as tidying a desk: visible progress, immediate dopamine, no real risk. So you keep doing it. Every new container ships with someone's pre-baked dashboard, you import it, and now you have eighteen panels for a thing you check once a month.
The dashboards I actually look at, it turns out, number three. One overview that tells me if anything is on fire, one for the storage pool because that is the thing most likely to ruin my weekend, and one for the network because DNS is always the answer eventually. The other twenty-eight are archaeology.
So I am doing the unglamorous thing and deleting most of them. Not archiving, deleting. If I genuinely need a panel again I can rebuild it in five minutes, and that small cost is exactly the filter I want. A monitoring setup should answer one question fast: is everything fine, and if not, what. Thirty-one dashboards answer a different question, which is whether I have too much time on my hands, and the answer to that one is clearly yes.