Operations
From the field archive · Learned in 2020
Anatomy of a Weekend Dynamics 365 Go-Live
The cutover runbook I helped a staffing firm through in June 2020, generalized into a checklist, and updated for backups, administration mode and pipelines in 2026.
In June 2020 I spent a weekend on calls with a staffing firm moving its recruiting teams onto Dynamics 365. The plan fit on one page: bring systems down Friday night, take backups, lock things down, license users, smoke test, keep users out until Sunday afternoon, and start hypercare Monday morning. Most of it went to plan. The parts that didn't are the reason I still use a version of that page with customers today.
The building blocks are the same, but the details moved. Administration mode now has clearer documented behavior for background operations and Power Automate flows. System and manual backups are kept for seven days by default, and production Managed Environments can extend retention to 28 days. Mailbox approval can be delegated to a Dynamics admin. Data Export Service reached end of support in November 2022, and Microsoft now points to Azure Synapse Link for Dataverse or Microsoft Fabric link. Pipelines in Power Platform can deploy your final solution with validated connections and environment variables, instead of a manual import at 10 p.m.
01The shape of the weekend
A weekend cutover isn't about doing a lot of work. It's about doing a small amount of work in the right order, with clear decision points. Here's the structure that firm used, generalized:
Freeze and prepare
- Put the environment in administration mode
- Take manual backups
- Lock down connected legacy systems
- Start licensing users (independent task)
- Bring up briefly for a first smoke test, then close again
Check and fix
- Morning technical checkpoint
- Status note to the executive sponsor
- Optional afternoon checkpoint, decided in the morning
- Enable and fix mailboxes
Open and support
- Bring systems up mid-afternoon
- Full smoke test, then a checkpoint call
- "System is up" note to the sponsor
- Hypercare starts first thing Monday
Two choices in that plan were smart. Licensing ran in parallel with everything else, because it didn't depend on anything. And the Saturday afternoon checkpoint was optional, decided on the morning call. Don't schedule meetings you might not need. Do schedule the decision about whether you need them.
02Lock it down: administration mode
Administration mode lets only users with the System Administrator or System Customizer role sign in to a sandbox, production, or trial (subscription-based) environment. That's what you want during a cutover: admins working, users out.
The optional setting matters just as much. Turning off background operations disables Dataverse asynchronous operations such as workflows and Exchange synchronization. Email isn't sent, and server-side sync of appointments, contacts and tasks stops. Here's what still runs, because it surprises people:
- Plug-ins and custom workflow assemblies still run on the Dataverse platform.
- Scheduled (recurrence) Power Automate cloud flows still run. They live in the Power Automate service, not in Dataverse asynchronous operations.
- Flows with a Dataverse trigger, such as "when a row is added, modified, or deleted," don't fire while background operations are off.
List every scheduled flow and integration that writes to the environment, and decide which ones to turn off for the weekend. Administration mode won't stop them for you. Also note that System Customizers must open the environment directly by URL while it's in administration mode.
03Backups: know your retention first
The backup step is the one I'd underline. On the Thursday before that go-live, the customer asked whether we could restore one of their environments from a backup older than seven days. They had taken a manual backup earlier and hadn't expected it to age out. It had.
That retention still applies, with one useful change:
| Backup fact (Oct 2026) | What it means for your go-live |
|---|---|
| 7-day default | System and manual backups for production and sandbox environments are kept for seven days by default. |
| Up to 28 days | For production Managed Environments, a tenant admin can set retention to 7, 14, 21 or 28 days. The setting applies to system and manual backups. |
| No default environment manual backups | Manual backups work for production and sandbox environments, but not the default environment. |
| 1 GB free to restore | You need at least 1 GB of free capacity to restore. If you're over capacity, plan for that before the weekend, not during it. |
| No offline copy | You can't download a database backup for offline use. If you need a long-term copy, plan another approach. |
| Restores take time | Copying and restoring data can take more than a day for large environments, especially with audit data. |
The lesson from that Thursday wasn't technical. Nobody had written down when the backup would expire. Today I ask for a fresh manual backup right before cutover, labeled with the go-live name, and I put its expiry date in the runbook next to the rollback decision deadline.
04Licensing and mailboxes in parallel
Licensing users can start on Friday night alongside everything else. Once users are licensed, the ones who need Outlook integration can have their mailboxes enabled. In practice, mailbox enablement is where Saturday goes.
On Sunday afternoon that weekend, a list of several dozen mailboxes that wouldn't enable arrived just before the checkpoint call. The support ticket was already open, which helped. We reran Test & Enable for one user first, confirmed what changed, and then worked through the rest. My email archive post covers the three gates every mailbox goes through and why batches fail together.
- Set server-side synchronization as the default for new users before you license anyone.
- Assign the Delegated Mailbox Approver role ahead of time, so you aren't hunting for a Global or Exchange admin on Saturday.
- Enable a pilot group of mailboxes on Friday night. Enable the rest only after the pilot works.
05Checkpoints and one thread
The first thing I asked for once the cutover started was simple: one email thread with every participant on it, so we stayed in constant contact. It sounds obvious. But in most cutovers I've seen, the customer team, the partner, and Microsoft each have their own thread, and decisions get made in one and missed in the others.
- One thread, one bridge. Everyone on it, including whoever owns DNS, Exchange and integrations.
- Checkpoint calls with a fixed agenda. What's done, what's blocked, go or no-go for the next step.
- A short sponsor update after each checkpoint. Two lines: on schedule or not, and what you need from them.
- A named decision owner for rollback, with a deadline that's earlier than your backup expiry.
06Hypercare and support readiness
Hypercare started at 7 a.m. Monday. The next week brought the expected questions about access and views, and one that wasn't really a go-live issue: how to feed Dynamics data to the analytics team. In 2020 that meant Data Export Service to a data lake design. Today it means Azure Synapse Link for Dataverse or Fabric link. Either way, it belongs in the go-live plan, because reporting teams notice missing data on Monday.
Before the weekend, make sure the people who can open support cases know how, have the environment and tenant details ready, and know your support plan's severity definitions. When something breaks at 9 p.m. Saturday, nobody should be looking up how to file a ticket.
07What I'd add today
Pipelines in Power Platform
Deploy the final solution with connections, connection references and environment variables configured up front. Deployments are prevalidated against the target, and pipelines support scheduled deployments.
Managed Environments
Required for the 28-day backup retention on production. Also unlocks governance features covered in my governance toolbox post.
Capacity check
Restores, copies and new environments are blocked without enough free capacity. Check the capacity page the week before. My storage archive post covers the usual culprits.
08The runbook checklist
| When | Step | Owner |
|---|---|---|
| T-2 weeks | Confirm backup retention, capacity headroom, support contacts, and DNS/Exchange owners | Platform lead |
| T-1 week | Dry run in a sandbox copy, including restore; list flows and integrations to pause | Platform + integration leads |
| T-1 week | Assign Delegated Mailbox Approver; set server-side sync defaults | Dynamics admin |
| Fri evening | Administration mode (background operations off if needed); manual backup; record expiry | Dynamics admin |
| Fri evening | Deploy final solution through a pipeline; lock down legacy systems; start licensing | Release lead |
| Sat | Checkpoint calls; mailbox pilot then batch; sponsor update | Cutover lead |
| Sun | Leave administration mode; full smoke test; go or rollback decision; "system up" note | Cutover lead + sponsor |
| Mon onward | Hypercare desk; daily triage; analytics feeds confirmed | Support lead |
Where I'd start
Write the runbook on one page, with an owner and a time for every step. Then check three things that bit people in 2020: when your backups expire, who can approve mailboxes, and which flows and integrations keep running in administration mode. Do a full dry run, including a restore, in a sandbox. A calm go-live weekend is mostly decided the week before.
Happy to go deeper on any of this.
— K.G
Sources
- Microsoft Learn: Administration mode
- Microsoft Learn: Back up and restore environments
- Microsoft Learn: Dataverse capacity-based storage overview
- Microsoft Learn: Connect to Exchange Online
- Microsoft Learn: Overview of pipelines in Power Platform; Run pipelines in Power Platform
- Microsoft Learn: Transition from legacy data integration services to Fabric link and Azure Synapse Link for Dataverse
- Microsoft Power Platform Blog: Do more with data – From Data Export Service to Azure Synapse Link for Dataverse (November 2021)
Get new posts by email
Provider: your email provider
