The Weekend Padlock: Assigning Custody of TLS Certificates in UK Cloud Estates
Photo: Saishna96, CC BY-SA 4.0, via Wikimedia Commons
A browser warning is a poor notification system. By the time a customer sees a padlock turn red, the organisation has usually known for weeks that a certificate was due to expire, but nobody has owned the renewal as a single, accountable task. The certificate itself is not complex. The difficulty lies in the estate around it: load balancers, content delivery networks, API gateways, SaaS custom domains, internal services and machine-to-machine endpoints, each with a different renewal path and a different team nearby.
For UK organisations running mixed cloud estates, certificate expiry is rarely a failure of technical skill. It is a failure of custody. The certificate has a location, an issuer and an expiry date, but no named person who can say, without checking, when it will be renewed and who will verify the deployment.
Why expiry so often lands on a Sunday
Short-lived certificates have improved security, but they have also shortened the distance between a missed renewal and an outage. A ninety-day certificate renewed by an automated process can still fail on a Sunday if the automation was configured by a contractor, the deployment credentials have expired, or the alert goes to a shared mailbox that is not monitored at weekends.
Weekend expiry is also a scheduling problem. Renewals are changes. If they are left to the last permitted day, they arrive outside normal change windows, often when the fewest people are available to validate a chain, check a load balancer configuration or roll back a failed deployment. The certificate may be valid, but the organisation has made the renewal harder than it needed to be.
Begin with a register, not a scanner
A scanner will tell you what it can see. It will not tell you who owns the service behind a certificate, whether the certificate is deployed in more than one region, or whether a renewal will require a change request. A useful certificate register is therefore a custody record first and a technical inventory second.
Start with a controlled register that records the certificate, the service it protects and the person accountable for its renewal. Scanning can populate and verify the register, but it should not be the register itself. The distinction matters because scanners miss private certificate authorities, internal mTLS endpoints and SaaS integrations where the certificate is managed by a supplier but still affects a UK business service.
Scope the estate in four layers
A practical scope for most UK cloud estates has four layers. The edge layer includes public load balancers, CDN certificates and web application firewalls. The application layer includes API gateways, ingress controllers and service meshes. The SaaS layer includes custom domains, integration endpoints and any certificate used for SAML signing or secure mail flow. The internal layer includes private certificate authorities, internal service certificates and client certificates used for machine-to-machine authentication.
Each layer needs a different discovery method. Edge certificates may be visible from the public internet; internal certificates may only be found through configuration management or certificate authority logs. The register should state which method was used, so that a gap in coverage is visible rather than assumed away.
The fields that make a register useful
A register with too few fields becomes a list of expiry dates. A register with too many becomes a data entry exercise that nobody maintains. The useful middle ground includes the certificate subject and alternative names, the issuer, the certificate type, the expiry date, the deployment location, the business service it supports, the renewal method, the automation status, the named custodian and the escalation route.
Two further fields are often overlooked: what depends on the certificate and what happens if renewal fails. A certificate on a customer-facing payment page carries a different risk from one on an internal reporting tool. Recording that difference allows the organisation to prioritise verification effort and to decide which renewals need a rehearsed fallback.
One accountable custodian, not a committee
Certificate ownership often falls between teams because each team holds part of the process. The web team requests the certificate, infrastructure deploys it, security sets the policy, and procurement pays the certificate authority. When the expiry date approaches, each team can reasonably assume another is monitoring it.
A custody model resolves this by naming one accountable custodian for each certificate, even if several teams contribute. The custodian does not have to perform every step. They must be able to confirm that the renewal is scheduled, that the deployment is verified and that the old certificate has been retired. For suppliers and SaaS integrations, the custodian is the person inside the UK organisation who maintains the relationship and confirms the supplier's renewal evidence.
Automate the check before automating the renewal
Automation is valuable, but it should not be treated as a substitute for custody. A daily expiry check is a low-cost control that gives the organisation time to act. Thresholds at thirty, fourteen and seven days are common; the alert should route to the custodian and to a backup, not only to a shared inbox. The alert should also be tested. An expiry warning that nobody has seen in a test is an assumption, not a control.
Where automated renewal is appropriate, it still needs an owner, an exception process and a post-renewal verification step. Automated deployment can fail silently if permissions have changed or if the new certificate is not picked up by every instance. The register should record the last successful verification, not just the last successful issuance.
Treat expiry as a scheduled change
A certificate renewal is a change event, and it should appear in the change calendar alongside other production changes. That simple discipline reduces the chance of a Sunday surprise. Renewals can be scheduled during business hours, with a named verifier and a rollback position. Change freezes around Christmas and other peak periods should be reflected in the renewal plan, not discovered during them.
A quarterly rehearsal adds further assurance. Select a sample of certificates, confirm that the alert fires, check that the custodian can access the renewal path, and verify that the fallback certificate or rollback procedure is still valid. The rehearsal does not need to be elaborate. It needs to answer one question: would the organisation know before the browser does?
The padlock is a promise
A TLS certificate is a promise that the service is the service it claims to be. That promise is renewed by people and processes, not by the padlock icon. A register with clear custody, tested alerts and scheduled renewals turns certificate expiry from a weekend incident into ordinary operational hygiene. For UK cloud estates, that is a modest control with a disproportionately large effect.