Ways to visualize error counts over time and drill down by text matches in logs (e.g. “bulk delete)?
We need a single graph showing total error count over time with the ability to break it down by any ad-hoc search (e.g. error stack or message contains "bulk delete"). What’s the best approach/architecture to: create a stable timeseries metric from logs/errors that match arbitrary text queries, allow fast drilldown/splitting by that query (and other fields) in dashboards? Any notes on cost/cardinality and best practises for indexing/structured fields would be helpful.
2 Replies
Hello,
I think that our other conversation is your best bet but you could also create an
error table and populate it from the captured errors in a boot plugin. Then using Gelly you could create a view that can be queried in the frontend of your app.
It would take a long time to explain everything in detail but I'd be glad to talk through specific issues you run into down the line; that is, if you choose to take this route.
Here are some docs that might help!
https://docs.gadget.dev/reference/gelly
The one thing that you wouldn't have here is a search since you can't search (semantic/elasticSearch) on a computed value but you could filter to see error logs for an exact string matchThanks, Antoine! That makes sense — creating an error table with a boot plugin and exposing it through a Gelly view sounds like a solid approach. I’ll explore that setup. Appreciate the pointers and the clarification about the payload and search limitations.