Notifications and alerts
Get told when a deploy fails, a backup does not run, a server goes offline, or a token gets minted.
Getting told when something happens: a deploy failed, a backup did not run, a server went offline, somebody minted an API token.
How it works
Two halves that are configured separately.
A channel is somewhere to send a message. There are twelve kinds and you can have as many of each as you like.
An alert selection belongs to each channel, not to the team. A team chat room that wants every deployment outcome and an on-call phone that wants only failures is the normal case, so each channel carries its own list of the 35 alerts it cares about.
A channel nobody has configured sits on the catalogue defaults, so a team gets something sensible before it ever opens the page.
Channels
Settings -> Notifications, then add a channel.
| Kind | What it needs |
|---|---|
| Discord, Slack, Mattermost, Microsoft Teams, Lark, Webhook | An incoming webhook URL |
| Telegram | A bot token and a chat id |
| Your own SMTP server, or a Resend API key | |
| ntfy, Gotify, Pushover | The service's URL and token. Beta |
| Browser push | Permission from your browser. Beta |
Each channel has a test, so you find out now rather than during an incident. Credentials are write-only: leaving a field blank when you save keeps what is already stored.
Pick what each channel is told about
Open the channel's alert picker. It is the same search box and category browse as the role editor, so it should feel familiar. The 35 alerts group into nine categories:
| Category | Examples |
|---|---|
| Deployments | Deployment failed, Deployment succeeded, Deployment interrupted, Git connection stopped working |
| Apps | App keeps restarting |
| Cron jobs | Cron job failed, Cron job finished |
| Databases | Database ready, Database setup failed, Database rebuilt, Database deleted |
| Backups & restore | Backup failed, Backup finished, Restore finished, Restore failed |
| Servers | Server offline, Server back online, Server running hot, Server disk almost full, Server cannot run apps, Server identity changed, Server agent certificate not renewed, Cleanup failed, Leftover containers |
| This Deplo instance | A new Deplo version is available |
| Security & team | API token created, API token revoked, Member joined, Member removed, Member access changed, Team owner changed, Two-factor requirement changed, Repeated failed sign-ins |
| Domains & TLS | Certificate expiring, Domain points elsewhere |
Unselect all and Reset to defaults are both there, spelled the way you expect.
Alerts are not the activity trail
Two different things, on purpose.
- An alert interrupts. It goes to a channel, now, because somebody should look.
- The activity trail is looked up later. It answers "who did this and when", in the Activity page, and it is never delivered anywhere.
Limits and gotchas
- Every webhook URL must be public HTTPS. An outbound guard refuses addresses inside your own network, which means a Gotify or ntfy on the LAN is refused. That guard is what stops a notification URL being used to probe your internal network.
- Every alert key here has a real emitter. Nothing in the list is a switch that promises an alert and delivers silence.
- Notification settings are per team. Another team's failures go to their channels.
manage_notificationsis the capability.
If it does not work
- The test succeeds and real alerts never arrive: that channel's alert selection does not include them. Open its picker.
- Saving a channel is refused: the URL is not public HTTPS, or it resolves to a private address.
- Email never arrives: check the SMTP host, port and credentials, and whether your provider requires a verified sender address.
See also
Did this page help you?