fly-logs · diff
git:20260711.696597d to git:20260922.a4bd47b
2 added, 2 removed. Audit A to A.
What this update changes in the checks
No check changed: every fixed check gives the same answer before and after this update.
A tripped check is a label, not a verdict: read the matched text and the lines below. This week's updates that newly trip a check.
---
name: fly-logs
description: >-
- Fetch and filter production logs from Fly.io apps. Trigger on "посмотри
- логи", "check prod logs", "что в логах", or when debugging any production
+ Fetch and filter production logs from Fly.io apps. Trigger on "show me the logs",
+ "check prod logs", "what is in the logs", or when debugging any production
issue on a Fly-hosted app.
---
# Fly logs
## Fetch
```bash
fly apps list # find the app name
fly logs -a <app> # live tail
fly logs -a <app> --no-tail | tail -200 # recent burst, non-blocking
fly logs -a <app> -i <machine-id> # one machine only
```
`fly logs` tails by default; for analysis prefer `--no-tail` piped through
grep so the command terminates.
## Filter
```bash
fly logs -a <app> --no-tail | grep -iE 'error|traceback|exception' | tail -50
fly logs -a <app> --no-tail | grep ' 500 ' # server errors on requests
fly logs -a <app> --no-tail | grep '<endpoint>' # one route's traffic
```
## Read the result properly
- A traceback's LAST frame in app code (not library code) is where to look.
- Repeated identical errors every N seconds = a crash-looping background job
or health check, not user traffic.
- Shape-mismatch / feature-names errors on prediction routes = FEATURE_COLS
drift between serving and training repos; fix the sync, don't restart.
- For deeper analysis hand the evidence to the `debugger` subagent.