Running DNS for one organization is manageable, but **multi-tenant monitoring across a client portfolio** is a different job. Agencies, MSPs, and freelance sysadmins often look after 20-200 client domains spread across Cloudflare, Route 53, GoDaddy, and whatever registrar a client’s nephew picked in 2014. This article covers how to structure DNS monitoring so each client’s DNS records, subdomains, and email authentication are watched separately, alerts reach the right people, and nothing gets missed when a client leaves or a staff member moves on.
Why a Shared View Breaks Down at Scale
With three clients, a single dashboard and a shared inbox for alerts works fine. At thirty, it falls apart. Alerts for an e-commerce client’s checkout subdomain end up next to a typo warning on a dentist’s brochure site, and both get the same priority.
The main problem is **context**. A CNAME pointing at a deprovisioned Heroku app on one client’s domain is a subdomain takeover risk that needs fixing today. On another client’s domain it may be a staging environment that the client knowingly keeps. Without per-tenant ownership and baselines, the person on call can’t tell these apart quickly.
Responsibility is also split. In most agency setups, the agency manages some records and the client’s internal IT or a third-party vendor manages others. Monitoring has to show who owns each record, or every alert turns into an email chain about who touched what.
Structuring Multi-Tenant Monitoring for Each Client
An experienced MSP engineer builds the tenant model first and adds tools afterward. Each client becomes an isolated unit: its own domain list, discovered subdomains, baseline, alert routing, and contact list. A practical rollout looks like this:
1. Inventory every apex domain per client. Ask for registrar exports, not memory. Clients nearly always forget defensive registrations, country-code domains, and the microsite domain marketing bought for a 2022 campaign.
2. Run subdomain discovery per tenant. Use certificate transparency logs, passive DNS, and zone data where you have it. Expect 3-10 times more hostnames than the client’s own list.
3. Snapshot a baseline. Record current A, AAAA, CNAME, MX, TXT, NS, and CAA values before changing anything. That snapshot is how you tell drift apart from intended state.
4. Tag ownership and criticality. Mark each record as agency-managed, client-managed, or vendor-managed, and rate its business impact.
5. Define alert routing per tenant. Critical alerts go to the on-call engineer and the client’s named technical contact. Informational changes go into a weekly digest.
The practical setup for agencies monitoring multiple clients covers tooling choices in more depth. The point here is the order: tenant boundaries first, then checks.
A Scenario Most Agency Engineers Will Recognize
Take a mid-sized agency with about 60 retainer clients. A client ends a contract in March. The website moves to a new vendor, but nobody removes promo.clientdomain.com, a CNAME to a HubSpot landing page instance the agency had set up. Six weeks later the agency’s HubSpot portal is downgraded and the landing page domain is released.
The record still resolves. The client no longer pays the agency to watch it, and the new vendor doesn’t know it exists. This is how a dangling DNS record forms: two parties each think the other is responsible.
Per-tenant monitoring catches this in two places. The offboarding baseline flags every record pointing at agency-controlled services, and continuous checks flag the CNAME once its target stops serving the expected content. A clear client domain handover procedure closes the gap formally, with a signed list of records the client now owns.
Myth: One Global Alert Threshold Works for Everyone
A common belief is that a single standard rule set, such as “alert on any NS change, any MX change, any new subdomain,” can be applied to the whole portfolio. It sounds consistent. In practice it generates noise for some clients and leaves others under-protected.
A SaaS client deploying through Terraform may produce 40 legitimate record changes a week. A law firm might change one record a year. One rule set either buries the SaaS client in alerts or misses a meaningful change at the law firm. Thresholds need to be set per tenant, based on each client’s normal rate of change. Advice on alert thresholds that reduce false positive noise applies here, with the extra rule that every tenant gets its own baseline.
Some alerts should stay the same across all clients: nameserver changes, registrar lock removal, domains within 30 days of expiry, and CNAMEs pointing at services listed as vulnerable in the community-maintained “can-i-take-over-xyz” project. Those are never routine.
Email Authentication Across a Portfolio
SPF, DKIM, and DMARC are where multi-tenant DNS health problems pile up without anyone noticing. Clients add Mailchimp, Zendesk, and a new CRM, and each vendor asks for another SPF include. Under RFC 7208, an SPF record that needs more than 10 DNS lookups returns a permerror, and receivers may treat the mail as unauthenticated.
A seasoned deliverability lead checks lookup counts every month for every tenant, not only when a client complains about spam folders. DMARC policies vary too: some clients sit at p=none for years because nobody owns the move to quarantine. Tracking DMARC policy level per client gives you a concrete, billable improvement project.
Common Mistakes in Portfolio Monitoring
Using one API credential for everything. One Cloudflare global API key or one AWS IAM user with Route 53 access across every client is a single point of compromise. Use per-zone scoped tokens and per-account read-only roles. If one leaks, only one tenant is exposed.
Monitoring only what the agency manages. Engineers often watch the records they created and ignore the rest of the zone. Attackers don’t care who created a record. Discovery has to cover the whole domain, with ownership tags added afterward.
Treating offboarding as a sales event, not a technical one. When a contract ends, monitoring usually stops the same day. Keep read-only monitoring running for 30-60 days after exit, or until the handover checklist is signed, because that’s when dangling records appear.
Edge Cases Where the Standard Approach Changes
Clients in regulated sectors, such as healthcare, financial services, or entities covered by NIS2 in the EU, may contractually forbid third parties from holding API credentials. In those cases, monitor through external resolution only and send findings to the client’s own change process.
Clients with split-horizon DNS or on-prem Microsoft DNS servers expose only part of their namespace publicly. External monitoring will never see internal records, so the scope has to be written into the contract. Very small clients, such as a five-person office on Google Workspace with a single website, often need just quarterly reviews plus expiry and NS alerts. Applying enterprise-level monitoring to them adds cost without reducing much risk.
FAQ
How many clients can one engineer realistically monitor?
With per-tenant baselines and tuned thresholds, one engineer can usually triage alerts for 50-100 clients. Without tuning, the number drops quickly because most of the day goes to dismissing noise.
Should clients see their own monitoring dashboards?
Usually yes, read-only. Showing clients their own DNS health makes ownership clear and cuts the “why did you change our MX record” escalations, since the change history shows who made each change.
What happens to monitoring when a client switches agencies?
Keep monitoring for a defined handover window, export the baseline and change history to the client, and confirm in writing which records pointed at agency-owned services. Then delete the agency’s credentials and tenant data.
Summary
Multi-tenant monitoring comes down to separation: separate inventories, baselines, thresholds, credentials, and alert routes for each client. Keep a short list of alerts that are critical for every tenant, such as nameserver changes, registrar lock removal, expiring domains, and takeover-prone CNAMEs, and tune everything else to how each client actually operates. If you start with one thing, make it a baseline snapshot for every client before the next contract ends, since that’s the point where records are most likely to be left dangling.
