Skip to content

Status pages for your clients: a setup guide for agencies

Give each client their own status page, on a subdomain of their own domain, showing the services they use under the names they use. Point the subdomain with a CNAME to the status service, make sure a certificate can be issued for it, let the client's staff subscribe, announce maintenance in advance, decide who may post incidents, and test the whole page before you hand out the link.

A status page for a client works best when it belongs to that client: one page per client, on a subdomain of the client's own domain such as status.example.com, showing the services the client uses under the names the client uses. Technically, that means a CNAME record, a certificate for the subdomain and a subscriber list. Organisationally, it means deciding who posts incidents, how maintenance is announced and testing the page before the link goes out. This guide covers the setup; what to write on the page during an incident is a topic of its own.

One page per client or one shared page?

For an agency or a managed service provider, the default should be one page per client. The reasons are practical:

  • Each client sees only their own services. A shared page shows every client the names of the others, which is a confidentiality problem on its own.
  • The overall state means something. "Partial disruption" on a page with twenty clients tells a single client nothing about their shop.
  • Subscriptions fit. A client's staff subscribe to their page and are not woken up by another client's outage.
  • The page can carry the client's domain and language.

A shared page makes sense for one thing: infrastructure that all clients use in the same way, such as a hosting platform, a shared mail service or a central backup system. Many hosting providers run both – one platform page and one page per client for the client-specific services.

Which components should a client page show?

Think from the client's side. They do not know web-01 or the database cluster; they know "Online shop", "Checkout", "Customer portal" and "Contact form". Each component on the page should be one of those, backed by the checks that tell you whether it works for a visitor – usually an HTTPS check from outside, perhaps a check of a key page such as the checkout.

Keep it short. Three to six components per client are typical. If a component is not something the client would call you about, it does not need its own row.

How do you put the page on the client's domain?

Status services usually publish a page under their own address and let you add a custom domain. The client – or you, if you manage their DNS – creates a CNAME record for a subdomain:

status.example.com.   3600   IN   CNAME   status-pages.example.net.

Replace status-pages.example.net. with the target your status service gives you. Many services ask for a second record, usually TXT, that proves the domain belongs to your account and not to someone else who points a CNAME at the same service. Check the result before you go further:

dig +short status.example.com CNAME
dig +short status.example.com
status-pages.example.net.
status-pages.example.net.
203.0.113.20

The first command returns only the CNAME target, the second follows it to the address. If the first command prints nothing, the record is missing or not yet visible; ask the authoritative name server directly with dig +short status.example.com CNAME @ns1.example.com to rule out caching.

Two details cause most of the trouble:

  • No CNAME on the bare domain. example.com itself cannot carry a CNAME. Use a subdomain, or the A and AAAA records your service gives you – which the client then has to change if the service ever moves.
  • The name in the provider's interface. Most DNS interfaces append the zone automatically. Entering status.example.com there creates status.example.com.example.com. Enter status instead.

How does TLS work for the custom domain?

Visitors must reach https://status.example.com without a certificate warning, so a certificate for exactly that name must exist on the status service. Most services obtain it automatically once the DNS record points to them. When it fails, the cause is usually on the client's side:

  • An AAAA record that points elsewhere. Certificate authorities often try IPv6 first. An old AAAA record for the subdomain makes the validation fail even when the CNAME is correct.
  • CAA records. If the client's domain restricts which certificate authorities may issue certificates, the one your service uses must be allowed:
dig +short example.com CAA
0 issue "letsencrypt.org"

No output means no restriction. Once the page is live, check the certificate from outside:

openssl s_client -connect status.example.com:443 -servername status.example.com </dev/null 2>/dev/null | openssl x509 -noout -subject -enddate
subject=CN=status.example.com
notAfter=Apr 21 08:15:02 2027 GMT

A status page with an expired certificate is worse than none, so monitor it like any other certificate – see TLS certificate expiry.

Who gets notified, and how?

The people who need the page most are the client's staff: their support team, their shop manager, whoever answers the phone when the site is slow. Ask the client who that is and invite them to subscribe. A subscription should use a confirmation link, so nobody can sign up an address that is not theirs, and every message should carry a one-click unsubscribe.

Subscriber lists contain personal data. Agree with the client whose privacy notice applies and link it next to the form. This is general information, not legal advice.

How should maintenance be handled?

Announce it in advance when clients will notice: a reboot of the shop server, a database migration, a move to a new server. Give start, end, the affected components and the expected impact a few days ahead, so that subscribers hear about it before it starts. During the window, the affected components should show "maintenance" instead of "outage", and the on-call person should not be alerted for the expected downtime.

Keep it internal when clients will not notice: a service restart that takes seconds, work on a redundant node. Mute your own alerts for the window, but do not fill the client's page with notices that do not affect them – too many announcements teach subscribers to ignore them.

The rule of thumb: if a visitor could notice it, announce it.

Who may post incidents?

Decide this before the first incident, per client:

  • Your on-call technicians post and update incidents. At least two people per client should know how.
  • The account manager may want to read along, but should not be the only one who can post.
  • The client usually reads and subscribes. It should not have to report outages of the systems you run.

Agree with the client on a rhythm, for example an update every 30 minutes during an incident, and keep a short template ready:

Investigating – since 09:40 the online shop is responding slowly.
Checkout is affected; the website itself loads. We are working on it.
Next update by 10:15.

How do you test the page before going live?

  1. Open the page from a network outside your office, without being signed in.
  2. Read every component name as the client would. Remove anything that looks like a host name.
  3. Check DNS, certificate and redirects from http:// to https://.
  4. Subscribe with a test address of your own, confirm, unsubscribe.
  5. Announce a test maintenance in the future, check how it looks, cancel it.
  6. Stop one checked service briefly, or point a check at a wrong address, and watch the page turn red and back.
  7. Only then send the link to the client and put it in their website footer.

Common mistakes

  • Internal names on a client page. The page is public. A server name or a version number is information for attackers, not for clients.
  • One shared page for everyone. Clients see each other, and nobody understands what affects them.
  • DNS records with a wrong name. The zone appended twice, a missing verification record, a forgotten AAAA record.
  • Nobody subscribed. A page that no one at the client knows about does not reduce a single support call.
  • Maintenance announced only on the page. Without subscribers, nobody sees it before it starts.
  • No agreement on who posts. During the first incident, everyone assumes someone else does it.

Checklist

  1. One page per client; a shared page only for shared infrastructure.
  2. Three to six components named as the client names them.
  3. CNAME and verification record set, checked with dig.
  4. Certificate valid, no stray AAAA, CAA allows the issuer.
  5. The client's key people are subscribed.
  6. Maintenance with visible impact is announced days ahead.
  7. Two people per client can post incidents; a template is ready.
  8. The page is tested from outside before the link goes out.

How Lynx Heartbeat helps

With the agency plan, each client gets a sub-tenant with its own status page, which shows the monitors you choose and no server details, versions, ports or causes. The page can run on the client's own domain with two DNS records; Lynx checks them, obtains the certificate from Let's Encrypt and renews it. Owners and technicians can announce maintenance in advance and mute the alerts of the affected monitors for its duration, and visitors subscribe by email with a confirmation link. The state on the page follows the monitors automatically; Lynx does not write the explanation of an incident for you. More on the agencies page.

Frequently asked questions

Should an agency run one status page per client or one for all clients?

One per client, as soon as clients use different services. A shared page shows every client the names of the others and makes each of them search for their own row. A shared page only makes sense for infrastructure that all clients use in the same way, such as a hosting platform.

How do I put a status page on the client's own domain?

Create a CNAME record for a subdomain such as status.example.com that points to the host name your status service gives you, plus any verification record it asks for. Check it with dig +short status.example.com CNAME. A CNAME is not possible on the bare domain itself.

Why does the certificate for my custom status domain fail?

Usually because the DNS record does not point to the status service yet, because an AAAA record still points somewhere else, or because a CAA record on the client's domain does not allow the certificate authority the service uses.

Who should be allowed to post incidents on a client's status page?

The people who run the client's systems and are on call, at least two of them. The client itself usually reads the page and subscribes; it should not have to post updates about your work.

Websites

Related articles

All articles