Skip to content
Esc

Notifications

Notification channels for external services, the in-app notification feed and Activity page, and notices.

Dockmesh has three separate ways of telling you something:

MechanismWhere it showsWhat feeds it
Notification channelsSlack, email, ntfy and other external servicesAlert rules only
Notification feedThe bell in the top bar and Activity → NotificationsAlerts, failed backups, scans, image updates, agents going offline, migrations and more
NoticesA banner at the top of the app, or ActivityConditions that need someone to act

Channels are managed on Alerts → Channels. Click New channel, pick a type and fill in its fields. A channel receives messages only from the alert rules that select it, both when a rule fires and when it resolves. Backups, scans and other events don’t go to channels.

TypeConfig fields (required in bold)
webhook (Generic Webhook)url
ntfyurl (topic URL), token, priority (1–5; unset follows the level: 5 critical, 4 warning, 3 otherwise)
discordurl (webhook URL)
slackurl (incoming webhook URL)
teamsurl (webhook URL)
gotifyurl, token (application token)
email (Email (SMTP))host, port (default 587), username, password, from, to (list)
pagerdutyintegration_key (Events v2 routing key), client, client_url
pushoverapp_token, user_key, device, sound

The generic webhook POSTs JSON with title, body, level, container, metric, value, threshold and timestamp (Unix seconds). Use it to feed Telegram bots, n8n, Home Assistant and similar tools.

Each channel has a Language (English or German) for the messages it sends; channels without one send in English. Email is sent per recipient: an address that belongs to a Dockmesh user gets the mail in the language that user picked under My account, every other address gets the channel’s language. Recipients with different languages get separate mails.

Test on a channel’s row sends a test message. The Last delivery column shows when the channel last sent, or the error of the last failed delivery.

A failed delivery is retried twice (after 2 and 10 seconds) before Dockmesh gives up. HTTP channels time out after 10 seconds per attempt; SMTP has no timeout.

Each channel can be disabled, and it can carry a weekly quiet-hours schedule (mute_schedule, API only) during which it’s skipped:

json
{"type":"weekly","ranges":[{"days":[1,2,3,4,5],"from":"22:00","to":"06:00"}]}

Reading channels needs alerts.view, managing them needs alerts.create, alerts.update (also for Test) and alerts.delete, all from an assignment on everything. Secrets are never returned: tokens, passwords, keys and the webhook URLs of webhook, ntfy, discord, slack, teams and gotify come back empty with a <field>_set: true marker. Leave such a field empty when you edit the channel to keep the stored value.

The bell in the top bar shows the latest notifications with an unread count; Activity → Notifications has the full history, together with finished operations and open notices. You can mark entries read, mark all read, or delete them.

Entries come from these producers. Each entry only reaches users who hold the listed permission on the affected host or stack:

KindWhenWho gets it
alert.fireAn alert rule firesalerts.history on the host
backup.failA backup run failsbackups.view on the job’s host; system.update on everything for the system backup
backup.okA backup run finishes with warnings (clean runs send nothing)Same as backup.fail
cve.foundA vulnerability scan finds critical or high findingsimages.view on the scanned host
image.updateA running image on the server host has a newer version upstreamimages.view on the server host
agent.offlineAn agent disconnectshosts.view on that host
migration.complete, migration.failedA stack migration endsstacks.view on the stack
deploy.ok, deploy.failA git-triggered deploy endsstacks.view on the stack
system.upgradeA newer Dockmesh release is availablesystem.upgrade
invite.acceptedSomeone accepts an inviteThe user who created the invite

Entries aren’t pruned.

Notices are raised for conditions that need someone to act. Dockmesh raises one while the condition holds and removes it once it’s fixed. Each notice targets an audience: everyone, admins, specific users, or holders of a permission. Severity is info, warning or critical.

  • Only notices about something that stopped working are shown as a banner. You can hide one for 24 hours; it comes back while the problem lasts.
  • Warning and critical notices are listed under Needs attention on the overview and under Activity. A notice hidden from the banner moves there until it is resolved.
  • Info notices explain a change and are listed under Activity → Notifications. They can be dismissed; critical notices that aren’t in the banner stay until resolved.

Examples are legacy API tokens without an owner (see API tokens), proxy sites that aren’t served because of their upstream (see Reverse proxy) and agents too old to stream container events for alert rules (see Alerts).

Notices are checked at startup and whenever the condition changes. The API is GET /api/v1/notices and POST /api/v1/notices/{id}/dismiss.

See also Alerts and Monitoring.