Page MenuHomePhabricator

Kibana: Sort normalized messages by how many users they affect
Open, Needs TriagePublic

Description

While looking through the new client error dashboard for Wikidata, we noticed that we have a lot of errors that stem from the common.js of individual or a very low number of users. These are not what we are primarily looking for in this dashboard, and they might make it harder to find more important issues.

One way to cut down on that noise would be if we could filter/order the normalized messages by how many users they affect over the selected amount of time.

Would such a two-stage metric be possible inside Kibana/logstash?

Event Timeline

This should be possible with the unique count feature

I've added this table to the dashboard:

I've added this table to the dashboard:

Awesome, thank you!

I'm hitting similar issues. I've suggested tagging bugs when a Central Notice banner is displaying to help diagnose banner issues in T265096 and I've been actively thinking about a generic gadget vs user script tag in T262493 but we haven't reached any consensus yet on what that might look like so your thoughts are very welcome!

Arguably this is a better metric than error volume in most scenarios. Maybe we can use the unique IP count by default, e.g. for the graph across the top.