Microsoft's preliminary incident report says Exchange Online could not move off a problematic network gateway for scale reasons. What does that mean for continuity planning? Is a separate mail gateway worth it, and how do you make sure people can actually use it during an outage?
How do you plan for email continuity when Exchange Online can't fail over?
Answers
A continuity gateway can accept and queue incoming mail ahead of Exchange Online, and some services offer a separate web inbox while the main service is unavailable. That can be useful if a day without email has a real operational cost.
Check the details before buying, though. Queueing messages is not the same as giving users a working mailbox, and the gateway itself becomes another dependency. Compare its reliability and security with the outage risk you're trying to reduce.
Reply to this answer
The tool is only half the plan. If nobody remembers it exists or knows their login, it won't help much when mail goes down. Put activation steps in a runbook, assign more than one person who can communicate them, and test the process with users periodically.
Reply to this answer
Start with what actually breaks, not just whether staff can read their inboxes. Mail delivered to pagers and health monitoring alerts may be more urgent than ordinary correspondence. Those dependencies deserve their own outage plan and an alternative way to reach the right people.
Reply to this answer
A separate inbox won't necessarily fix the other end of a conversation. If customers and partners are affected by the same outage, they may not be able to receive or respond to mail either. Consider which communications can move to chat, phone, or another channel, and which genuinely need email continuity.
Reply to this answer
Don't rely solely on the affected mail service to tell you it's down. Use an external address for service health notifications or enable push notifications in the admin app, then make sure someone is responsible for watching them.
For queued mail, verify the provider's current retry and expiration behavior rather than assuming an incident-specific extension applies to every queue.
Reply to this answer
Add your answer
Answers are reviewed before publication.
_c84~nexlora.3:
Also check how users sign in. Single sign-on may be convenient, but it can be another dependency if the identity service has a problem too.