I want to inventory servers that can network boot, not just the ones currently using it. Is there a practical way to check remotely, and how do you verify that provisioning networks are actually isolated?
How can I audit which servers have PXE enabled?
Answers
Start with the management interfaces. Many servers expose BIOS and UEFI attributes through Redfish, so a script with appropriate credentials can query them without rebooting each host. Vendor management tools may also be able to read the settings and apply configuration profiles.
Check the attribute names for each model and firmware version. Network boot enablement, boot order and a one-time boot override can be separate settings. Record which machines you could not query as well as the results, so an incomplete inventory does not look like a clean one.
Reply to this answer
For hosts you can access through the OS, vendor utilities or firmware management interfaces may provide another way to collect settings. On Linux, efibootmgr can show UEFI boot entries and order, but that is not necessarily the same thing as a complete inventory of whether PXE is enabled in firmware.
Use those results alongside management-controller data rather than assuming a missing or low-priority network boot entry means the feature is disabled.
Reply to this answer
Keep the provisioning service on a separate deployment network, and move machines onto it when they need a build. Production networks should use authorized DHCP servers that do not advertise network boot services.
Then test the isolation, not just the diagram. Check which ports and VLANs can reach the deployment network and whether traffic can cross from other segments. Also restrict access to the management interfaces that can change boot settings.
c.57!marzipanhinge:
That makes sense. The network check seems useful even before the firmware inventory is complete.
Reply to this answer
DHCP logs are a useful shortcut for finding machines that actually ask for boot files. They won't tell you every machine with PXE enabled, though. A server may never have attempted a network boot during the period you checked.
Use logs as a supplement to the settings audit, and map observed MAC addresses back to the asset inventory. Avoid rebooting production servers just to generate requests unless that test has been planned and approved.
Reply to this answer
Exposure depends on more than a setting being enabled. Look at when a server might network boot, who can reach that network during boot, and who can control reboots or boot settings. A narrow opportunity may lower the risk, but it does not replace firmware patching or network isolation.
Reply to this answer
Add your answer
Answers are reviewed before publication.
Tivnexa.7r:
Some changes may require a reboot to take effect, even if you can read or submit them remotely. Check the vendor's behavior before scheduling a bulk change.