Reading your notifications

There is a bell in the top right of the panel. It holds what has happened to your hosting: an account that was created, a backup that failed, a disk filling up, a server that stopped reporting.

The bell is not a timeline of the whole platform. It shows exactly what you are allowed to see on the Audit log screen and nothing else. If you manage one account it is about your account; if you are a reseller it is about your own customers.

Unread, read and seen

The panel keeps track of what you have already read on the server, per person. That is what makes your laptop and your phone agree: mark something as read at your desk and it is read on the train too. Nothing needs refreshing — it catches up within a few seconds.

There are two different things here, and they really do differ:

The distinction is deliberate. Glancing at the bell on your phone is not the same as "I have dealt with this", and a panel that conflates the two is a panel that loses an important message to a single glance.

To start with a clean slate, use Mark everything as read at the top of the bell. It applies to everything you can see at that moment, on all your devices at once.

Eighteen failed tasks are one line

If something goes wrong eighteen times on the same server on the same day, you get one notification about it, with a counter: Task failed (18×). Click it and all eighteen are read — because those eighteen together are the one thing you have to do something about.

The panel groups on four things at once: the group the notification falls in, what it is about (usually the server), what happened, and the day. The day is in there on purpose: if you cleared it on Monday you want to be warned again on Tuesday, not have Tuesday quietly folded into Monday. From three like notifications upwards they collapse; two separate lines stay two separate lines, because you can simply read those.

The Notifications screen

The bell is a glance sideways. The whole picture is on Notifications, and it has three things the bell does not.

A tab per group. Along the top: All groups · Servers · Accounts · Security · Migrations · System, each with the number of unread notifications in it. One look tells you where something is waiting.

An urgency filter. Under that, five chips: Everything · Critical · Warning · Info · Done, each with the number of lines you will be left with if you press it. That way a filter can never empty your screen unannounced.

Both live in the address bar. /notifications?category=security&severity=crit is a link you can send a colleague, and it opens exactly the same view for them.

An expanding panel on the heavy ones. Anything critical, anything from security, anything from a migration, an advisor finding and a run of failed tasks all get a chevron on the right. Open it and you see:

• what to do with it — one sentence, written for that kind of notification;
• the context — when, which server, what it is about, how many times;
• several buttons to the screens where you can actually fix it (the server, its advice, its logs, the audit log).

Ordinary activity — "account created", "version promoted" — deliberately stays one line with one link. Two clicks to read that something went well is one too many.

Snoozing instead of dismissing

There is no "delete". There is snooze: a notification you can do nothing about right now goes away for an hour, a day or a week, and then comes back.

1. Open the bell, or go to Notifications in the menu.
2. Click the … to the right of the notification.
3. Choose Snooze: 1 hour, 1 day or 1 week.

The line disappears from the list until that moment. To see it in the meantime, press Show snoozed at the top: it is back, with the time it returns on it. Reading it also un-snoozes it.

Snooze is offered on the notifications that can interrupt you — critical and warning. Everything else you simply mark as read.

Why no delete button? Because a warning you can make disappear is a warning nobody can later show was ever there. Snoozing does exactly the same thing on screen and leaves the record intact.

When there is nothing

Then there is a green panel: Nothing needs attention — everything has been read. That is an explicit statement rather than an empty screen on purpose, because a list showing nothing because there is nothing looks exactly like a list that is broken.

With a tab or an urgency filter on, the green panel says nothing is open within this filter — and that another group or urgency will show you more.

What a notification says

Every notification answers three questions:

• What happened — in English or Dutch, depending on your language setting. Never a technical code.
• What it means — one line underneath, for instance why you are getting it.
• Where to go — click it and you land on the screen where you can do something about it: the account or the server.

To the left of the text is a colour: blue for a statement, green for something that went well, amber for a warning, red for something that is broken or about to be.

Only amber and red interrupt you with a pop-up. Info and "done" go to the bell and to the 08:00 daily summary — a panel that pops up for every job that went fine teaches you to dismiss pop-ups without reading them, and then the one that mattered gets dismissed too.

A pop-up with a button on it stays until you close it yourself. That is not forgetfulness: a timer that takes the button away while you are reading the sentence above it makes the button useless.

Notifications and the "Needs attention" panel

The dashboard has a Needs attention block. It looks like this list and it is something else, and the difference is exactly one word: now.

A server that went down at 04:12 and came back at 04:20 is gone from the dashboard and still in your notifications. Both are true: "it is down" no longer is, "it was down" still is.

How long notifications are kept

Ninety days, and the newest two hundred per person. The audit log is the full history of the platform and is kept far longer; the bell is the history of what you were told, and that does not need to scroll for ever.

Getting fewer notifications

To stop hearing about a whole group, go to Settings → Notifications — see Setting your notifications and e-mail. A group you switch off there leaves the bell alone but stays in your history: muting is not deleting.

For administrators: from the command line

On the panel server, corecp-panel inbox answers the questions that come with a support ticket. Reading and snoozing deliberately are not there: those are the person's own decisions.

# Which kinds of notification exist, and how much do they weigh?
corecp-panel inbox kinds
# KIND                 CATEGORY  SEVERITY  COPY FROM  TITLE
# health.node.down     fleet     crit      health     A node is unreachable
# quota.disk           account   warn      mail       Disk space nearly full
# task.failed          system    warn      inbox      Task failed

# "I was never told" — what would this person have seen?
corecp-panel inbox show --person anna@test100.nl
# anna@test100.nl — 3 unread, 1 unseen, 1 critical (read through 2026-08-14T09:12:00Z)
# WHEN                  SEV   CATEGORY  KIND         COUNT  UNREAD  LINK
# 2026-08-14T10:02:00Z  warn  system    task.failed  18     true    /nodes/stck1.corecp.dev