Operate
Alerts
Each agent watches its own host and tells the dashboard when a resource runs hot or an agent goes quiet. You set the thresholds; ConnXL raises, escalates, and resolves the alerts and can push them to Slack, Teams, Discord, or your own webhook.
6 min read
What's monitored
Every agent instance reports its host health — CPU, memory, and disk. Each metric has two thresholds:
- Warning — default
85%. The resource is getting tight. - Critical — default
95%. Act now.
Change them per agent from the thresholds card on the agent runtime page, or disable a metric you don't care about. Thresholds are evaluated on the agent with a little hysteresis, so an alert doesn't flap when a value hovers right on the line, and a new setting applies live over the config channel — no restart.
The alert lifecycle
- Raised — a metric crosses its warning threshold. It shows on the agent's alert banner, the health column, and the topbar bell.
- Escalated — a warning-level alert crosses critical.
- Resolved — the metric drops back below warning. The alert closes on its own.
- Agent offline — if an agent stops heart-beating for more than 10 minutes it's flagged offline; it clears automatically the moment the agent reconnects.
Notification channels
Alerts live in the dashboard by default. To get them pushed somewhere, add a notification channel under your organization's webhooks. Five kinds are supported — slack, teams, discord, custom (any HTTPS endpoint), and email — and each channel subscribes to the events it cares about:
- Alert events — raised, escalated, and resolved.
- Agent status — an agent going offline or coming back online.
- Update available — an agent is running an older build than the latest published release (once per agent and release).
- Function failures — a batch of failing calls, with the function, a count, and a sample error.
Custom webhooks are signed and guarded
A custom channel's payloads are HMAC-signed so your receiver can verify they came from ConnXL, and the destination is checked to keep webhooks from reaching private network addresses. Use Send test to fire a sample event while you wire it up.
Email channels write in each recipient's language
An email channel has no URL — it sends to a list of organization members you pick (leave it empty and it goes to every admin). Each person receives the alert in their own language, and each can mute alert emails for themselves under Organization → My preferences. See Members & roles for how that language is chosen.
Logs & exports
Every function call an agent runs is reported back as a usage event — never the data, only the metadata. The dashboard rolls those events into analytics you can read on screen or download as a spreadsheet. This page also covers the separate agent process-log viewer used to diagnose a broken function.
Troubleshooting
Task-oriented fixes for the problems you'll actually hit: a blank formula, an Excel error code, an agent that won't reconnect, and enrollment failures — plus where to look for more detail.