Last week, I wrote about the Microsoft 365 outage and what it means for digital sovereignty. One of our clients didn’t lose a single course registration, because failed emails went into a queue instead of disappearing.
Email is one dependency. Login is another, and for many organizations it’s the same provider. If your editors sign in to TYPO3 with their Microsoft account, Entra ID decides who gets into your CMS. So what happens when Entra ID isn’t reachable?
As far as I can tell, sign-in wasn’t the main problem this time. Reports differ on how much Entra ID was affected. But identity providers do fail, and it’s better to answer this question before it happens than during it.
What Single Sign-On Actually Depends On
With Single Sign-On, TYPO3 doesn’t check passwords anymore. When an editor logs in, TYPO3 sends them to the identity provider. The provider checks who they are, including the second factor, and sends a signed confirmation back. TYPO3 trusts that confirmation and starts a session.
That has a few consequences:
- New logins need the identity provider. If Entra ID, Keycloak, or ADFS can’t be reached, nobody gets a new confirmation, and nobody can log in through SSO.
- Running sessions keep going. The confirmation is checked once, at login. Editors who are already logged in can keep working until their TYPO3 session ends.
- Your website keeps running. Visitors who don’t log in don’t notice anything.
What We Learned From Last Week
The outage made us look at our own Single Sign-On module with one question in mind: what exactly happens, step by step, when the identity provider can’t be reached? We used the answers to fine-tune a few things.
TYPO3 needs the identity provider’s metadata: its addresses and the certificate it signs with. Because of last week, we made one thing configurable: how long TYPO3 keeps using the identity provider’s last known metadata when it can’t fetch a new copy. The default is 24 hours. A longer time bridges a longer outage. A shorter one means TYPO3 stops trusting an old certificate sooner. Every organization can decide that trade-off for itself.
How long editors and visitors stay logged in is a separate setting, and it’s in TYPO3 itself: the session timeout for the backend and the frontend. Once someone is logged in, TYPO3 doesn’t ask the identity provider again, so that timeout also decides how long people can keep working during an outage.
Two more changes make the module sturdier when the provider is gone:
- The metadata is cached. TYPO3 fetches it with a three-second timeout, reuses it for an hour, and waits a minute before trying again after a failure. A slow or unreachable provider no longer slows down or breaks other requests. Before, an unreachable metadata URL could lead to error pages in the backend and on websites with SSO.
- Without metadata, SSO is unavailable, and nothing else is. Logged-in editors keep working, the password login stays available unless you enforce SSO, and websites that don’t require a login look as usual. A website that does require SSO answers with a “503 Service Unavailable” page instead of its content. It stays closed, because the identity provider being down is no reason to open it.
Both changes will be part of version 2.2.0. If you run our module and configure your identity provider by URL, please update.
For me, that’s the real lesson of last week: don’t wait for an outage to find out how your systems behave in one. Walk through it while everything still works.
“Then Why Use SSO at All?”
That’s a fair question after a week like this one. Without SSO, TYPO3 keeps its own user list and its own passwords. If Microsoft is down, your editors can still log in.
But that comes with its own problems, and they show up more often than outages:
- People leave, and their TYPO3 accounts stay. The directory entry is disabled the same day. The TYPO3 user isn’t, because nobody remembers it exists.
- Your password rules and second factor stop at the CMS. Everything your IT department enforces in the identity provider doesn’t reach TYPO3.
- Agencies keep logins long after projects end. Unless someone removes them by hand, they keep working for years.
They rarely make the news, but they come up in security audits. I still think SSO is the right choice for most organizations with a central directory. You just need a plan for the day the provider isn’t there.
How Our Single Sign-On Handles an Outage
With the two fixes in place, this is what our module does when the identity provider is down:
- It keeps a copy of the provider’s metadata. As described above: up to 24 hours by default, configurable, with a short timeout for every new attempt.
- There’s always a way back in. The TYPO3 Install Tool has its own password and doesn’t depend on SSO. If the provider stays away, an administrator can log in there and switch SSO off for the backend or the frontend. That only works if someone has the Install Tool password, so check that before you need it.
- Password login stays unless you turn it off. The SSO button sits next to the normal login. If you keep local accounts for a few administrators and don’t enforce SSO, they can still log in when the provider is down.
- Backend and frontend are configured separately. You can use SSO for one and not the other, or different providers for each. If both use the same provider, an outage affects both, of course.
- You can see what’s going on. Rejected SSO logins are written to the TYPO3 system log, and every failed attempt to fetch the metadata is logged as an error.
- There’s no b13 service in between. The module runs inside your TYPO3 instance. Your login depends on your identity provider and your server, not on us.
What It Can’t Do
No SSO module can log you in through Entra ID while Entra ID is down. That’s how SSO works. What a module can do is make sure the outage stays a login problem and that you can decide what happens next.
A few more limits you should know about:
- The cached copy includes the provider’s signing certificate. If you replace that certificate, or if it’s compromised, flush the TYPO3 caches right away so TYPO3 fetches the new metadata. That’s also why the cached copy has a time limit by default.
- Disabling a user at the identity provider stops their next login, not their current TYPO3 session. If access has to end immediately, disable the user in TYPO3 as well.
- Our module only speaks SAML 2.0 for now, not OpenID Connect. Entra ID, Keycloak, and ADFS all support SAML, so in practice that’s rarely a blocker.
Is Ours Better?
There are several SSO extensions for TYPO3, open source and commercial, for SAML, OpenID Connect, Shibboleth, and LDAP. Many of them do the basic job well. And as you’ve read above, we improved ours this week, too.
Where I think our module stands out is in everything around the login: the documented way back in, the cached metadata, the separate backend and frontend setups, the visible errors, and the fact that new users aren’t created automatically unless you decide they should be. Those are the things that matter on a bad day, and they’re the questions I’d ask about any SSO setup, ours included.
Six Questions to Ask Before the Next Outage
- How do we get into TYPO3 if the identity provider is down? Who has the Install Tool password, and where is it stored?
- What happens to our website if the provider’s metadata can’t be fetched? Does the SSO setup fetch it on every request, and is that failure handled? Test it on staging with an address that doesn’t resolve.
- Do we have emergency accounts? Microsoft itself recommends emergency access accounts in Entra ID that don’t depend on the normal sign-in path. The same idea works for TYPO3: one or two local admin accounts with strong passwords, documented and tested.
- Does our password reset depend on the same provider? If editors get their reset link by email and your email runs through Microsoft 365, that path is down too. This is where the two posts meet.
- Are backend and frontend independent? Would a problem with editor login also lock out your customers or members?
- Have we tested it? Switch SSO off on staging once and see how long it takes to get back in. Better to find out on a quiet Tuesday than during an outage.
SSO is still the right choice for most organizations we work with. Just make sure someone knows what to do when the login page stops working.
Not sure what happens to your TYPO3 login when Entra ID is down?
Let’s go through your setup together: identity provider, emergency access, and what your editors and visitors would see.

