A hurricane-season IT checklist for a Houston office

• Which functions have to run through an outage: mail, files, sign-in, phones, the order or warehouse system; and which can wait for power, like the print server and the badge readers
• Where people work when the building is dark: home, another site, a hotel lobby; and whether their laptop, their sign-in, and their files work from there
• Who declares that the outage plan is in effect, and how staff hear it when the mail server is off
• Which vendors get called first: the ISP, the generator service, the copier dealer, the landlord, and where their account numbers are kept

• A UPS sized for a clean shutdown, not for running the business: enough minutes for the hosts to shut their virtual machines down in order, and a network card in the UPS so it can tell them to
• A shutdown script that has actually been triggered from the UPS, in the right order: virtual machines, then the host, then the storage, with the firewall and the switch last
• A generator transfer switch that the generator's own service company has tested under load, with the date of the last test written on the panel
• The power path labelled, so the person on site at 6am knows which strip feeds which host
• An LTE or 5G failover on the firewall, because the fibre often survives when the power does not, and sometimes it is the other way round
• Equipment off the floor and away from windows, which is the one item on this list that is about water

• The 3-2-1 rule, with the off-site copy immutable, so a storm and a ransomware note are covered by the same copy
• Microsoft 365 backed up separately: mail, SharePoint, OneDrive, and Teams, because retention is not a backup, and a deleted site is gone when the retention window closes
• The restore rehearsed with a stopwatch and the time written on the plan, so the recovery estimate is a measurement
• The backup console's login in the company password manager and known to two people, not in one head that may be evacuating

A company whose sign-in, mail, and files live in Microsoft 365 has an office that can go dark without the business going dark: people sign in from wherever has power. A company whose only directory is a domain controller in the closet has laptops that keep working on cached credentials for a while, and no way to reset a password or add a device until the closet comes back. Password hash sync to Entra ID is what turns the second company into the first, and it takes an afternoon; the hybrid or cloud-only decision is a longer one.

Check the multi-factor methods too: a phone that drowned in a truck cannot approve a sign-in, so everyone with an administrator role needs a second method registered, and the break-glass account's codes need to be sealed somewhere dry.

• A phone system that fails over to mobiles, or a published mobile number that does, with the change rehearsed once outside a storm
• A laptop-goes-home rule that starts when a storm is named, not when it is in the Gulf
• A printed contact sheet: staff, vendors, the ISP account numbers, the generator service, the insurer
• An out-of-office and a website notice ready to paste, because the person who would write them will be busy

1. Power path first: check the UPS batteries survived the discharge and the generator has fuel
2. Network, then the domain controller, then storage, then the application servers, in that order, and nothing else until sign-in works
3. Confirm the backup jobs resumed and the off-site copy is current
4. Walk the racks for anything that did not come back, and write down what the plan got wrong while it is fresh

Winter Storm Uri, in February 2021, was the grid failing rather than the sea rising: over 4.5 million Texas customers were without power, some of them for four days in freezing weather, from Houston to Dallas–Fort Worth. Every item on this list applies, and the season it belongs to is all year. The Texas section of our Enterprise IT services page has the figures, and the servers, virtualization, and backup work is where the closet items get done.