Server Management S2 — Cheatsheet
Deployment, GPO, AD infrastructure, file sharing, DHCP/DNS and recovery — 2024–2025
Table of Contents
- Base Installations
- PXE Deployment
- Group Policy
- Global Catalog and FSMO
- IAM Hardening
- NTFS and AGDLP
- SMB, DFS and NFS
- Redundant DHCP
- Redundant DNS
- Business Continuity, Backup and Recovery
Base Installations
An integration lab starts with consistent naming, a network plan, DNS, time and roles. Windows domain members use the AD DNS server. Linux hosts get a static address or reservation, the correct hostname/FQDN and time synchronization.
Test-NetConnection dc01.lab.local -Port 53
Add-Computer -DomainName lab.local -Credential LAB\Administrator -Restart
DNS before domain join
Many Kerberos and join issues are DNS issues. Test A/AAAA, SRV and reverse resolution before suspecting credentials.
PXE Deployment
PXE lets a client boot over the network and fetch an image/installation environment.
client DHCP/PXE discovery
→ IP + boot server/boot information
→ bootloader via TFTP
→ WinPE
→ installation image and unattended configuration
WDS or a deployment platform manages boot and install images. Watch out for BIOS/UEFI architecture, DHCP/PXE relay, VLANs, TFTP/firewall and driver support.
When DHCP and PXE run on different servers, prefer IP helpers/relay configuration. Blindly setting vendor options 66/67 can break different client architectures.
Group Policy
GPOs contain Computer Configuration and User Configuration. Processing typically follows LSDOU: Local → Site → Domain → OU; within the same level, link order determines priority.
| Mechanism | Effect |
|---|---|
| Enforced | prevents being overridden by a lower OU |
| Block inheritance | blocks non-enforced parent links |
| Security filtering | applies only to selected principals |
| WMI filter | applies based on a system property |
| Loopback | user policy follows computer location (merge/replace) |
gpupdate /force
gpresult /h C:\Temp\gpresult.html
Get-GPResultantSetOfPolicy -ReportType Html -Path C:\Temp\rsop.html
GPO Preferences configure settings but can be changed by the user/app; Policies are typically enforced. Don't put passwords in Group Policy Preferences.
Global Catalog and FSMO
The Global Catalog contains a full copy of the own domain and a partial attribute set of other forest domains. It supports forest-wide searches, UPN logon and universal group membership.
FSMO roles
| Role | Scope | Task |
|---|---|---|
| Schema Master | forest | schema changes |
| Domain Naming Master | forest | add/remove domains |
| RID Master | domain | RID pools for SIDs |
| PDC Emulator | domain | time, password priority, legacy/PDC tasks |
| Infrastructure Master | domain | cross-domain object references |
netdom query fsmo
Move-ADDirectoryServerOperationMasterRole -Identity DC02 -OperationMasterRole 0,1,2,3,4
A transfer happens in a controlled way with a working old DC; seize only when the old role holder definitively won't return.
IAM Hardening
- Separate regular and administrative accounts.
- Use least privilege and administrative tiering.
- Limit Domain Admins; use delegated administration.
- Set strong password/account-lockout/Kerberos policies.
- Use LAPS for unique local administrator passwords.
- Disable legacy/problematic protocols where possible.
- Audit privileged group changes and failed logins.
NTFS and AGDLP
Effective access combines share and NTFS permissions; over the network, the most restrictive combination applies. NTFS supports inheritance and granular ACLs.
Example:
Use Modify for normal data management; Full Control also allows changing permissions/ownership. Avoid individual ACEs and explicit deny unless truly necessary.
SMB, DFS and NFS
Windows SMB
New-Item -ItemType Directory D:\Shares\Directie
New-SmbShare -Name Directie -Path D:\Shares\Directie -ChangeAccess LAB\DL_DirectieShare_RW
Get-SmbShareAccess Directie
Get-SmbSession
Get-SmbOpenFile
SMB signing protects integrity/authenticity; SMB encryption also protects confidentiality. Don't use SMBv1.
DFS
- DFS Namespace provides one logical path, e.g.
\\lab.local\Data. - DFS Replication synchronizes folders between servers.
- A namespace is not automatic data redundancy; that requires multiple folder targets and replication.
Watch out for conflicts, staging quota, AD replication, backlog, and the fact that open files aren't replicated like a database cluster.
Samba
Samba provides SMB for Linux and can work standalone or AD-integrated. Both filesystem permissions/ACLs and share configuration must allow access.
NFS
# /etc/exports
/srv/data 192.168.10.0/24(rw,sync,root_squash)
exportfs -ra
showmount -e server
mount -t nfs server:/srv/data /mnt/data
NFS uses Unix identities; consistent UID/GID or centralized identity mapping is crucial.
Redundant DHCP
Windows DHCP failover replicates leases and configuration between two partners.
| Mode | Behavior |
|---|---|
| Load balance | both servers share clients |
| Hot standby | active server + reserve |
MCLT limits lease renewal during communication problems. Failover doesn't replace correct relay configuration or a backup of the DHCP configuration.
In hybrid DHCP, Windows and Linux can serve scopes, but avoid overlapping pools and keep options, reservations and DNS updates consistent.
Redundant DNS
AD-integrated zones use multimaster AD replication and secure dynamic updates. Linux/BIND can be a secondary for a classic primary zone via zone transfer, if explicitly allowed.
| Zone type | Property |
|---|---|
| Primary | writable master copy |
| Secondary | read-only copy via AXFR/IXFR |
| Stub | mainly SOA/NS/glue for delegation |
| AD-integrated | writable on DCs within replication scope |
Get-DnsServerZone
Get-DnsServerResourceRecord -ZoneName lab.local
Resolve-DnsName -Server dc02 lab.local -Type SOA
A client with two DNS servers doesn't use the second as round-robin load balancing, but per resolver behavior when there's no/an invalid answer. Both must correctly know the same internal namespace.
Business Continuity, Backup and Recovery
| Term | Meaning |
|---|---|
| RPO | maximum acceptable data loss in time |
| RTO | maximum acceptable recovery time |
| Backup | recovery copy of data/system state |
| HA | limiting downtime through redundancy |
| DR | recovery after a large-scale disaster |
3-2-1 rule
Modern extension: at least one offline/immutable copy and zero errors after verification.
| Backup | Contents |
|---|---|
| Full | all selected data |
| Incremental | changes since the last backup of any type |
| Differential | changes since the last full |
| System State | critical Windows/AD configuration |
Restore is the real test
A successful backup job doesn't prove RPO/RTO are achievable. Run periodic restore tests and document the full procedure, credentials and dependencies.
Recovery order follows dependencies: platform/storage → identity/DNS → network services → applications → data → functional validation.