Skip to main content
Careers · Interviews

Systems administrator interview questions, and what each one is really testing

A sysadmin interview in a Microsoft environment checks that you understand how Active Directory, Group Policy, Entra ID, Microsoft 365 and Intune behave when something breaks, and that you change live systems carefully. Learn the reasoning behind each answer rather than the wording, because the panel will follow up on whatever you say.

10 min read Sources checked 17 November 2026

A systems administrator interview in a Microsoft environment checks two things: that you understand how Active Directory, Group Policy, Entra ID, Microsoft 365 and Intune behave when something breaks, and that you change live systems carefully. The questions below are typical of junior and mid-level roles. For each one you get what the interviewer is listening for and what a good answer covers, with the Microsoft documentation it rests on. Learn the reasoning, not the wording: a panel will follow up on whatever you say.

Expect three kinds of question: concepts ("what does the PDC emulator do?"), faults ("a policy isn't applying") and your own record ("tell me about a change that went wrong"). For faults, talk through the order you would check things in and why, what you would expect to see at each step, and when you would stop and ask someone.

Active Directory and the DNS underneath it

"What are the FSMO roles, and what happens if one is unavailable?"

This checks that you know Active Directory is multi-master, so most changes can be written on any domain controller, but that a few kinds of change are handled by a single DC to prevent conflicts. Microsoft's article on FSMO roles lists five: schema master, domain naming master, RID master, PDC emulator and infrastructure master. There is one schema master per forest, and the domain naming master is the only DC that can add or remove domains in the forest. The RID master, PDC emulator and infrastructure master work at domain level.

A good answer moves on to impact, because that is the real question. The PDC emulator is the authoritative time source for its domain (the one in the forest root domain should take its time from an external source), password changes made on other DCs are replicated to it preferentially, and account lockout is processed there. Kerberos depends on that shared time, and the default maximum clock difference it tolerates is five minutes. The RID master hands each DC a pool of relative IDs, which the DC uses to build the SIDs of new users and groups, so a short outage goes unnoticed until a DC runs low and asks for more. The schema master only matters when someone extends the schema. Saying that roles can be transferred to another DC earns credit.

KBY's runbook on Kerberos logons failing because of clock skew shows what the time dependency looks like when it breaks.

"How does replication work, and how would you check it?"

The interviewer wants to hear that changes made on one DC reach the others, and that you know where to look when they don't. The Knowledge Consistency Checker (KCC) runs on every DC and builds the replication connections. Within a site it arranges writable DCs in a two-way ring; between sites, you define site links and the KCC creates connections between bridgehead servers (replication concepts).

For health, name the tool and the log. Microsoft's replication troubleshooting guide recommends checking status daily with repadmin /showrepl (repadmin /showrepl * /csv produces a spreadsheet covering every DC) and reading the Directory Service event log. It gives the usual root causes as network connectivity, name resolution, authentication, the directory database, the replication engine and a topology that doesn't match the real network, and singles out DNS misconfiguration as a common one. A strong answer also mentions the tombstone lifetime: a DC that has not replicated for longer than that is a serious problem, not a quick fix. KBY's replication failure runbook walks through one case.

"Why does Active Directory need DNS?"

Microsoft's DNS and AD DS page says AD DS uses DNS so that clients can locate domain controllers and DCs can talk to each other. Domain controllers register SRV records in DNS. A client looking for one queries records such as _ldap._tcp.contoso.com, then checks the DCs it finds with an LDAP ping (how DC location works). So a domain-joined machine pointed at a DNS server that doesn't hold the domain's records can't find a DC, and sign-in, Group Policy and anything else that needs a DC fails with it. Say you would use nslookup to check the records exist and point at the right servers.

Group Policy, and the policy that won't apply

"In what order is Group Policy processed?"

Local policy first, then GPOs linked to the site, then the domain, then organisational units, parent OU before child. Later policy overwrites earlier policy, so the GPO closest to the object wins a conflict (Group Policy hierarchy). Add that computer settings apply at startup and user settings at sign-in, with background refreshes after that (Group Policy overview), and you have a complete answer. A stronger one names what changes the default outcome: enforced links and blocked inheritance alter which GPOs a container receives, and security filtering narrows who a GPO applies to.

"A GPO isn't applying to one person's PC. Walk me through it."

A good answer moves from the cheapest check to the most involved:

  1. Confirm the GPO is linked where the object actually lives. Computer settings follow the computer object's OU and user settings follow the user's.
  2. On the PC, run gpresult /r for a summary, or gpresult /h report.html from an elevated prompt for the full report (gpresult). It shows which GPOs applied and which didn't.
  3. Run gpupdate /force and look again.
  4. Read the System log and the Group Policy operational log. Microsoft's Group Policy troubleshooting guidance notes that the operational log lists applied GPOs and denied GPOs with the reasons for each denial.
  5. If the client can't reach a DC properly, look at DNS and the network. The same guidance ties event 1129 to blocked LDAP on port 389, and event 1058 to the client failing to read the policy's files from SYSVOL.

Security filtering is a favourite follow-up. KBY's config trap on removing Authenticated Users from a security filter explains how a tidy-looking change stops policy reaching computers.

Hybrid identity, Microsoft 365 and Intune

"How do on-premises accounts get into Entra ID, and how do users sign in?"

Synchronisation keeps on-premises users and groups matched in Entra ID, through Entra Connect Sync or Entra Cloud Sync, which uses a lightweight agent on-premises. A good answer then names the three sign-in methods on Microsoft's hybrid authentication page. Password hash synchronisation sends a hash of the user's password to Entra ID and runs every two minutes. Pass-through authentication uses agents on servers on-premises to check the password directly against Active Directory. Federation hands sign-in to a separate trusted system such as AD FS.

The follow-up is usually a fault. "A user changed their password on-premises and can't sign in to Outlook with the new one" checks whether you think about sync timing. The same page notes two traps worth naming. With password hash sync, a change to an on-premises account's state, such as disabling it, isn't enforced in the cloud until the next sync. And a password set with "user must change password at next logon" isn't synced until the user actually changes it.

"A team leader needs to reset passwords for their team. What access do you give them?"

This checks least privilege. Microsoft's guidance on Microsoft 365 admin roles is to assign the role with the fewest permissions that does the job, such as Password Administrator or Helpdesk Administrator rather than Global Administrator, to keep Global Administrators as few as possible, and to require MFA for every admin. Add that you would want the request approved and recorded.

"What does Intune do, and what is a compliance policy?"

Intune is Microsoft's cloud service for managing devices and apps. A good answer separates mobile device management, where the device is enrolled and Intune manages the whole of it, from mobile application management, where Intune protects only the work apps and their data, which suits personal phones. Compliance policies are rules a device must meet, such as a minimum OS version. A noncompliant device can be marked, locked or sent an email, and Conditional Access can use the result to block access.

One detail that stands out: the tenant-wide setting "Mark devices with no compliance policy assigned as" defaults to Compliant. Microsoft advises changing it to Not compliant if you use Conditional Access, so that only devices confirmed as compliant get in.

Patching, backups and restores

"How would you roll out updates without breaking the business?"

The interviewer wants to hear about rings, testing and a way back. Microsoft's servicing strategy for Windows updates describes testing on a few devices first, recruiting pilot users, testing the most business-critical applications, then deploying broadly in rings, with each department split across several rings so one bad update can't stop a whole department. Name the tool you have used (WSUS, Configuration Manager or Windows Update client policies), say how you would pause a ring that is going wrong, and for servers mention the maintenance window and change record.

"How do you know the backups work?"

Only a restore proves it. A good answer covers testing restores regularly, keeping more than one copy and keeping at least one copy offline. The NCSC's guidance on offline backups explains why: it has seen ransomware encrypt connected USB and network drives holding backups, and connected cloud storage too. It describes the 3-2-1 rule (at least three copies, on two devices, one offsite) and adds that at least one backup should be offline at any given time.

The Active Directory version of this question is "someone deleted a user yesterday; how do you get it back?". If the AD Recycle Bin is enabled, the account can be restored with all its attributes, including its group memberships. It is not enabled by default, and once it is enabled it can't be turned off. If it was never enabled, getting the account back with its group memberships means restoring from a backup, and a sensible junior says they would involve a senior colleague before restoring anything into Active Directory.

Permissions, PowerShell and the Linux server

"A user can open a shared folder but can't save to it. Why?"

Two sets of permissions apply to a file share. Microsoft's file server guidance says share permissions and NTFS permissions are independent, and the more restrictive of the two applies. Share permissions apply only over the network; NTFS permissions apply locally and remotely. So Full Control on the share means nothing if NTFS gives the user Read. A good answer checks both, plus the user's group memberships, before changing anything. The same page mentions that some administrators set the share permission wide open and control access with NTFS alone.

"What do you use PowerShell for?"

Panels want an example, not a list of cmdlets. Reporting is the safest place to start: Get-ADUser -Filter * -SearchBase "OU=Finance,DC=contoso,DC=com" lists the users in one OU, and because Get-ADUser returns a default set of properties, you add -Properties for anything else. For a script that changes things, mention running it with -WhatIf first, which describes what the command would do instead of doing it, then testing on one object and logging what changed. KBY's script library has read-only examples worth reading before the interview.

"A service on a Linux server has stopped. What do you check?"

If the role includes Linux servers, expect something like this. Start with systemctl status nginx, which shows the unit's runtime state followed by its most recent log lines (systemctl), then journalctl -u nginx for its full log (journalctl). After that come the usual suspects: a full disk, a configuration change, a dependency that has stopped. Asking what changed recently is the answer the panel is waiting for.

The change that went wrong

Sysadmin panels often ask some version of "tell me about a change you made that caused a problem". It checks honesty, whether you rolled back quickly, how you told people and what you changed afterwards. Pick a real example where the mistake was yours. Walk through what the change was, how you noticed it had gone wrong, what you did first, how you kept people informed, and what is different now: a pilot group, a written backout plan, a second pair of eyes. Stories where the fault turns out to be someone else's do you no favours. KBY's guide to the STAR method covers how to build these answers.

"How do you feel about change management?" checks whether you see it as paperwork or protection. A good answer says that a change request makes you write down the risk, how you tested, how you will back out and who needs to know, and that an emergency fix still gets recorded afterwards. If your last job had no formal process, say what you did instead, such as warning the team before a change and noting exactly what you did in the ticket.

Getting ready for the day

Build the parts you are weakest on in a lab: a domain controller, a second DC so you can watch replication, a few GPOs to break with security filtering, and a Microsoft 365 test tenant if you can get access to one. KBY's identity and access lessons and certification paths help if you want a structured route. Then practise saying your troubleshooting order out loud. If the role also covers Azure, read the cloud engineer questions as well.

Courses and practice tests

Partner links
  • Amazon · PaidExam Ref MD-102 Microsoft Endpoint Administrator (2nd edition)Microsoft Press's exam reference for MD-102 Endpoint Administrator, for the Intune and Windows management side of a sysadmin role.Check partner offer
  • MeasureUp · PaidMeasureUp practice testsTimed practice tests for Microsoft role-based exams such as MD-102, once you have worked through the free Microsoft Learn material.Check partner offer
  • Pluralsight · PaidPluralsightVideo courses and labs on Windows Server, Active Directory, PowerShell and Microsoft 365 administration.Check partner offer

Partner links are sponsored: KBY may receive a commission at no extra cost to you. Whether a provider pays KBY does not decide what we recommend, and the free official material comes first. Advertising disclosure.

Common questions

What if I don't know the answer to a technical question?
Say so, then say how you would find out: which log you would read, which documentation you would check and who you would ask. Reasoning aloud from what you do know is worth far more to a panel than a confident guess that turns out to be wrong.
How is a sysadmin interview different from a service desk interview?
Service desk questions are mostly about fixing one person's problem and handling the person. Sysadmin questions are more about shared systems that many people depend on, so expect more on how things work underneath, how you test a change and how you would undo it.

Sources

Every figure and rule in this guide comes from these pages, read on 17 November 2026. Pay rates, exams and schemes change: check the source before you rely on one.

Spotted something wrong or out of date? Tell us and we will correct it.

All careers guides