← All posts

The Day Microsoft 365 Email Went Dark at Work

Around noon today, email at work just stopped. No new messages, no errors worth reading, nothing. At first I assumed it was something on our side. Then the group chats lit up, and it became clear this wasn't us. Microsoft 365 was having a very bad day.

What Actually Happened

Microsoft said it was investigating problems reaching Exchange Online mailboxes, starting in the Americas and then widening to basically everywhere. Outlook took the biggest hit, but Teams, SharePoint, Copilot and a few security tools were wobbly too. Outage trackers lit up with well over a hundred thousand reports.

The explanation so far is pretty mundane. Microsoft said a chunk of its infrastructure was under heavy load, not because of a traffic spike, but because a number of servers had unexpectedly dropped out and the rest were left carrying the extra weight. Their fix was to shift connections around to other parts of their global network, and to look at rolling back some recent configuration changes.

As I write this, there's still no official root cause, and no sign it was an attack. Just a reminder that even the biggest cloud on the planet can fall over because of a few servers failing health checks.

How It Hit Me

Email is one of those things you don't notice until it's gone. No messages since noon means no approvals, no replies, no "can you look at this?" and no way to know what I'm missing. It's weirdly quiet, and the quiet is the problem. Anything urgent that would normally land in my inbox is stuck somewhere I can't see.

It was highly disruptive. A good chunk of the day's work moves through email, and the workarounds (phone calls, chat, waiting) are all slower.

Our Backup Plan, and Where It Fell Short

We're not completely unprepared. We have a secondary domain and our own email server, so in theory we can keep sending and receiving when Microsoft is down.

In practice it only helps so much. Our customers don't know about the backup. They write to the addresses they've always used, and those live on Office 365. So when 365 is down, their messages go to the very place we can't reach. A standby server is great for outbound mail and for people who already know the alternative address, but it can't catch mail meant for a mailbox that's hosted somewhere else.

Mail sent to us during the outage should generally be queued and retried by the sender's server, so I'm hopeful most of it will turn up once things recover. But "hopeful" isn't a great place to be when a customer is waiting on an answer.

What I'm Taking From This

Redundancy has to include the part your customers can see. A backup server is only half a plan if everyone's still pointed at the primary. Here's my short list for next time:

  • Keep a clear, tested way to tell customers an alternative contact exists, even if it's just a phone number or status page that doesn't depend on Microsoft.
  • Know in advance which people need the backup address, and make sure they've actually used it.
  • Don't put the incident comms on the same platform that's broken.
  • Check your MX setup and whether a mail-queueing or filtering layer in front of 365 could buy you breathing room.

None of this makes outages painless. But the lesson I keep relearning is that convenience and single points of failure tend to come as a package. Today was a good, annoying reminder to go look at ours.