What it is for
This is the diagnosis screen: one error, everything known about it, and nothing you can change on it. Changing status, merging and writing a note live under Administering, and the split is deliberate: diagnosing and administering are jobs in opposite states of mind, and having both on one screen made everyone change things before understanding them.
What is on it
- The reading band: class, the whole message, status, environment and where it breaks. This is what you read standing up.
- The numbers: how many times, how many customers, how many versions, first and last seen.
- The 30-day curve: it sits next to the numbers and not six hundred pixels below, because it is the thing that says whether it got worse.
- The stack: from the most recent occurrence. If the project has a source map on file, it comes translated back to source.
- The debut deploy: the version this error first showed up in, when there are deploys on record.
- The timeline: what already happened to this error. Was it fixed before? Did it come back? Who touched it?
- The evidence: screenshots and attachments that came with the event.
What usually comes before it
When the application sends usage events, the screen lists the actions that showed up in the ten minutes before this error's recent occurrences. For each one it says before how many occurrences it appeared, how many times it happened in 30 days, and how many of every 1,000 were followed by this error. An action that comes before every occurrence but rarely breaks is just common, like opening the home screen; the one with the high rate is the suspect.
Who keeps hitting it
A map of the customers this error reached in the last eight weeks, one row per customer and one column per week, colored by the number of occurrences. A hundred occurrences at one customer that never stops is a different problem from a hundred spread across a hundred customers once each, and the total does not tell them apart.
What you can do from here
- Markdown: downloads the error as plain text, to paste into an issue or a ticket. Copying from the screen brings the dashboard HTML along.
- See the occurrences: takes you to the list already filtered by this fingerprint.
- Open the administration: to change things.
If this error is merged into another, the screen says so and points at the destination group. The occurrences count there, not here.
The address
The URL is /erro?fp=<fingerprint>. It is stable: paste it in
chat or in a ticket, and it still opens the same error months later.