Crash Alert
When a node dies unexpectedly, DendROS prints an inline banner immediately after the death line and repeats it periodically so you don't lose it in fast-scrolling output.
Or set it directly in ~/.config/dendROS/defaults.yaml:
crash_alert: true
crash_alert_color: node # node | red
crash_alert_interval: 30 # seconds; 0 = only on new crashes
What it looks like
The banner appears directly in the terminal after the death line — no separate window, no scroll regions. The terminal is completely clean after Ctrl-C.
Options
crash_alert_color
Controls the color used for node names inside the banner.
| Value | Effect |
|---|---|
node (default) |
Each node name is shown in its configured group color |
red |
All node names are shown in bold red, regardless of group |
node is useful when you have many groups and want to immediately know which subsystem crashed. red gives a uniform alarm-style look.
crash_alert_interval
Seconds between periodic banner reprints (default: 30).
| Value | Effect |
|---|---|
30 (default) |
Reprint every 30 s |
0 |
Print once per new crash event; no periodic reprints |
Configuring via dendros config
All three settings are exposed in the TUI under Crash alert:
| Setting | Values |
|---|---|
| Crash alert | on / off |
| Crash alert color | node / red |
| Crash alert interval | integer (seconds) |
Notes
- Traceback suppression — if a node printed a Python traceback before dying, the crash alert for that node is skipped. The traceback already communicates the failure clearly enough; a redundant banner adds noise.
- Ctrl-C suppression — once
Ctrl-Cis received, all subsequent node deaths are treated as expected cascade shutdowns and produce no banner. Only true mid-run crashes (before anyCtrl-C) trigger the alert. - Exit codes are shown next to the node name when available (
exit code 1,exit code -11). - The
[node-N]numeric suffix is stripped before the name is displayed in the banner. - If a node has no configured group color (unmatched node), the banner falls back to bold red.