Outlook outage, 2026-11-12
What happened#
On Thursday afternoon, US time, Microsoft said it was investigating reports of people having trouble reaching their mailboxes.
Getting to Outlook takes two steps. First your computer looks up where Outlook lives (DNS). Then it connects to one of the addresses it was given and asks for the page.
The first step worked: lookups returned addresses on 795 of 797 checks. Overnight, the second step mostly failed: most of the addresses they returned did not answer.
Who was affected#
Microsoft said the problem could reach people in any region (ABC News).
From our three probes (test machines in Phoenix, South Carolina and Amsterdam), Thursday 21:49 to Friday 07:07 UTC:
- Phoenix and South Carolina: 24 of 1,682 tries reached Outlook. About 1 in 70.
- Amsterdam: 177 of 834 tries reached Outlook. About 1 in 5.
Timeline#
tries that reached Outlook, all three probes, one cell per 30 minutes: every one 9 in 10 or more half or more under half none no readings
- to 21:36 UTC no readings we were not measuring yet
- start Microsoft says it is investigating reports of people having trouble reaching their mailboxes (ABC News)
- our first reading: DNS answers
- our sampler starts: 4 of 46 tries reach Outlook by 22:00
- to 11:35 UTC no readings the machine running our sampler restarted at 07:10, and the sampler was started again at 11:35
- coming back: 12 of 17 tries reach Outlook, up from 2 of 15 at 11:35
- fixed tries from all three probes reach Outlook again; 5 misses in the next 3,024
the readings, cell by cell
| UTC | tries that reached Outlook, all three probes |
|---|---|
| to 22:00 UTC | 4 of 46 |
| to 22:30 UTC | 8 of 108 |
| to 23:00 UTC | 18 of 110 |
| to 23:30 UTC | 23 of 86 |
| to 00:00 UTC | 17 of 67 |
| to 00:30 UTC | 17 of 132 |
| to 01:00 UTC | 7 of 144 |
| to 01:30 UTC | 6 of 144 |
| to 02:00 UTC | 18 of 168 |
| to 02:30 UTC | 5 of 144 |
| to 03:00 UTC | 10 of 144 |
| to 03:30 UTC | 11 of 144 |
| to 04:00 UTC | 5 of 143 |
| to 04:30 UTC | 7 of 144 |
| to 05:00 UTC | 12 of 168 |
| to 05:30 UTC | 8 of 144 |
| to 06:00 UTC | 6 of 144 |
| to 06:30 UTC | 11 of 144 |
| to 07:00 UTC | 7 of 144 |
| to 07:30 UTC | 1 of 48 |
| to 08:00 UTC | no reading |
| to 08:30 UTC | no reading |
| to 09:00 UTC | no reading |
| to 09:30 UTC | no reading |
| to 10:00 UTC | no reading |
| to 10:30 UTC | no reading |
| to 11:00 UTC | no reading |
| to 11:30 UTC | no reading |
| to 12:00 UTC | 54 of 99 |
| to 12:30 UTC | 120 of 167 |
| to 13:00 UTC | 121 of 144 |
| to 13:30 UTC | 126 of 144 |
| to 14:00 UTC | 143 of 144 |
| to 14:30 UTC | 144 of 144 |
| to 15:00 UTC | 168 of 168 |
| to 15:30 UTC | 144 of 144 |
| to 16:00 UTC | 143 of 144 |
| to 16:30 UTC | 168 of 168 |
| to 17:00 UTC | 144 of 144 |
| to 17:30 UTC | 144 of 144 |
| to 18:00 UTC | 143 of 144 |
| to 18:30 UTC | 168 of 168 |
| to 19:00 UTC | 144 of 144 |
| to 19:30 UTC | 144 of 144 |
| to 20:00 UTC | 143 of 144 |
| to 20:30 UTC | 168 of 168 |
| to 21:00 UTC | 144 of 144 |
| to 21:30 UTC | 144 of 144 |
| to 22:00 UTC | 144 of 144 |
| to 22:30 UTC | 144 of 144 |
| to 23:00 UTC | 167 of 168 |
| to 23:30 UTC | 144 of 144 |
| to 00:00 UTC | 24 of 24 |
What it looked like in DNS#
The three Outlook names we asked (outlook.office365.com, outlook.office.com, outlook.cloud.microsoft) pointed to one name, fs.outlook.mira.tm.svc.cloud.microsoft, and that name gave out between 1 and 8 addresses. The set changed from one ask to the next.
| what we asked | what came back |
|---|---|
| the Outlook names, four public DNS services, 21:36 UTC | an answer every time, with a valid signature (ad set) |
| the address set, three probes, Thu 21:49 to Fri 07:07 and 11:35 to 23:32 UTC | addresses on 795 of 797 asks; 8 addresses on 742 of them |
| the same question as if asked from five networks (RFC 7871) | 1 to 8 addresses, a different set per network |
| the delegation and the DNSSEC chain, 21:36 UTC | checked, no error |
Asks 1.1.1.1 for the address set that the three Outlook names end at.
dig -4 @1.1.1.1 fs.outlook.mira.tm.svc.cloud.microsoft AGO NOERROR, ad set, A records
NO GO SERVFAIL, NXDOMAIN, or no A record
From the Amsterdam probe, 21:40:06 UTC. The addresses, the TTL, the id and the Query time change from run to run.
Why#
Microsoft's words, as ABC News quoted them: the problem was "introduced by an issue during an earlier network capacity upgrade," and "This was a technical issue not related to cyber or AI / agents."
Our reading, 00:40 UTC Friday: the path from Phoenix and from Amsterdam to one silent address went into Microsoft's network and stopped there.
What we could not see:
- Anything before 21:36 UTC Thursday. The start time is Microsoft's, not ours.
- 07:07 to 11:35 UTC Friday: the machine running our sampler restarted at 07:10, and the sampler was not started again until 11:35.
- Inside Microsoft's network.
- IPv6, sign-in, and mail delivery.
- Your own network: test your own path.
If it happens to you#
- Check Microsoft's own status first: Microsoft 365 service health.
- Ask the name yourself: four public DNS services side by side.
Receipts#
Two kinds of reading, both from isitdns machines:
- The isitdns tools, 21:36 to 21:42 UTC Thursday: the Outlook names from the edge over DoH (Cloudflare, Google, Quad9, AdGuard) and 1.1.1.1 from the Phoenix and Amsterdam probes, plus the delegation, the DNSSEC chain and five client subnets.
- A sampler on three isitdns probes (Phoenix, South Carolina, Amsterdam), started by hand at 21:49 UTC Thursday, after the outage began.
- Each round it asked
1.1.1.1for the address set, then requestedhttps://outlook.office365.com/owa/once at each address: 5 seconds to connect, 8 seconds in all. - A try "reached Outlook" when
/owa/answered with its302redirect. A 500 or 503 does not count. - Its last round before the gap began at 07:06 UTC and ended at 07:07. The machine running it restarted at 07:10 UTC, and it was started again by hand at 11:35 UTC. Its last round was 23:32 UTC Friday.
- Each round it asked
Each try, by outcome:
| probe, window (UTC) | 302 | 500 or 503 | TCP, no HTTP answer | no TCP |
|---|---|---|---|---|
| Phoenix, Thu 21:49 to Fri 07:07 | 16 | 7 | 134 | 688 |
| South Carolina, same | 8 | 6 | 131 | 692 |
| Amsterdam, same | 177 | 29 | 255 | 373 |
| Phoenix, Fri 11:35 to 13:32 | 139 | 4 | 2 | 31 |
| South Carolina, same | 152 | 2 | 3 | 29 |
| Amsterdam, same | 130 | 1 | 11 | 50 |
| Phoenix, Fri 13:32 to 23:32 | 1006 | 1 | 0 | 1 |
| South Carolina, same | 1007 | 1 | 0 | 0 |
| Amsterdam, same | 1006 | 0 | 2 | 0 |
One try of each kind, curl -m 8 --connect-timeout 5 --resolve outlook.office365.com:443:<address> https://outlook.office365.com/owa/:
- 302: Amsterdam, 40.101.69.34, 00:38:04. TCP at 0.024 s, TLS at 0.114 s,
HTTP/2 302,location: https://outlook.office365.com/mail/ - 500: South Carolina, 40.99.165.114, 00:27:53. TCP at 4.18 s, TLS at 4.49 s, HTTP 500
- TCP, no HTTP answer: Phoenix, 52.96.164.178, 00:37:18. Connected at 4.127 s, client hello sent,
SSL connection timeoutat 5.0 s - no TCP: Phoenix, 52.96.163.34, 00:37:23.
connect to 52.96.163.34 port 443 failed: Connection timed outat 5.0 s
The path to the silent 52.96.163.34 entered Microsoft's network (AS8075) from Phoenix and from Amsterdam, and the last hop that answered was in 104.44.52.0/24 (AS8075) from both. The path to 40.101.69.34 answered end to end from both. mtr -T -P 443 from the probes, 00:40 UTC Friday.
The address set per client subnet, from Google with the subnet passed on (RFC 7871), 21:40:18 UTC:
| subnet | A records |
|---|---|
| 12.0.0.0/24 | 8, in 52.96.x and 52.97.x |
| 142.103.0.0/24 | 1, 52.96.164.146 |
| 200.160.0.0/24 | 1, 40.97.63.133 |
| 81.2.69.0/24 | 5, in 40.98.x, 40.101.x, 52.98.x and 40.99.x |
| 133.11.0.0/24 | 1, 40.99.76.114 |
Asked again at 21:42:20 UTC, the five read 1, 8, 8, 5 and 2.
Ask again:
| reading | api |
|---|---|
| the name, four resolvers | https://isitdns.net/api/query?name=fs.outlook.mira.tm.svc.cloud.microsoft&type=A&resolver=all&transport=doh |
| from Phoenix | https://isitdns.net/api/query?name=fs.outlook.mira.tm.svc.cloud.microsoft&type=A&resolver=cloudflare&transport=udp53&from=north-america |
| from Amsterdam | https://isitdns.net/api/query?name=fs.outlook.mira.tm.svc.cloud.microsoft&type=A&resolver=cloudflare&transport=udp53&from=europe |
| with a client subnet | https://isitdns.net/api/query?name=fs.outlook.mira.tm.svc.cloud.microsoft&type=A&resolver=google&transport=doh&dnssec=0&ecs=12.0.0.0%2F24 |