Patterns
Recurring shapes across logs and errors, and what they tend to precede.
Logs & Errors → Patterns looks for recurring shapes rather than individual events.
What a pattern is
A message shape that recurs with varying details:
connection to <host> refused after <n>msRather than ten thousand distinct lines, that is one pattern with a count, a trend, and a set of hosts it has applied to.
Why it is separate from categories
Log analytics categorises within one source. Patterns work across sources — the same shape appearing in application logs, browser errors and agent log watches is one pattern, and seeing it in three places at once is information that neither source has alone.
Trends
Each pattern carries a trend. The interesting ones are:
| Trend | Meaning |
|---|---|
| New | Has not been seen before. Nearly always worth a look |
| Rising | Getting more frequent |
| Stopped | Was regular, and has ceased |
"Stopped" is underrated. A pattern that fired every five minutes and stopped usually means the thing producing it stopped — which is rarely good news and never produces an error alert.
Using patterns
The practical workflow:
- Sort by new after a release. A new pattern is a new behaviour.
- Sort by rising when something feels slow but nothing has alerted.
- Check stopped when a downstream system has gone quiet.
Related
- Anomalies — statistical detection on metrics rather than shapes in text
- Error groups — thrown exceptions specifically
Where this behaviour lives: backend/internal/worker/log_category_worker.go, frontend/src/app/errors/patterns/. If the code and this page disagree, the code is right — please fix the page.
Part of Logs and errors — What the software said about itself, and what was unusual about it.