Centralized Fog server, remote appliances for PXE and host image deployment

[jaapvdpol]

I’am looking for a (partial) open source solution to transfer like 500 Windows 10 PC’s (with HP Secure Boot ) into a Linux OS (version not known yet).
These 500 PC’s are located in about 100 locations.
I already played in a local environment, with FOG and it worked great.
Besides the Secure boot issue I can transform a Windows PC to a Linux PC.
I use a NUC running Proxmox and a Debian VM with Fog on it.
Instead of managing 100 NUC’s with Fog, I was looking for a solution like running Foc on a central location, to manage the 100 NUC’s on the remote location.
These NUC’s are only for PXE boot and image storage/deployment.
I this sort of solution possible with Fog or are there other solutions?
Any advice is welcome!

Thanks in advance!!

[george1421]

@jaapvdpol In theory its possible to do with FOG. The concept in FOG terminology is master node at HQ and storage nodes at the remote locations. The master node will replicate the images created at HQ to all remote storage nodes. One caveat is that the remote locations must be in constant and routable contact with HQ 100% of the time. PXE booting computers at the remote site will contact the FOG master node (at HQ) to find its local storage node. Once identified it will image directly from that local storage node. Only status update packets are sent between the imaging remote computer and the FOG server at HQ. So by design it will work, I’m just not so sure about the scale you propose.

I wonder about your update frequency of these computers and if the proposed design is the right one based on how often you think you will reimage a remote computer. Will it be a one time imaging and then only reimage when the system fails? Would a mobile FOG deployment server work better for your deployment plan? In this design you would have a laptop running linux with FOG installed. You would also include dnsmasq with this mobile deployment server to provide pxe boot information to the pxe booting computers. You won’t need to adjust your remote sites networking to support pxe booting, you just drop the mobile deployment server on the network. Once booted you can pxe boot into the FOG iPXE menu and pick deploy image (system builder load and go methodology). You wouldn’t have post imaging management with the load and go method, but is that in your plan or only to migrate operating systems?

[jaapvdpol]

@george1421 Thank you for the answer.
Not sure what you mean about the scale I propose. You mean bandwidth issues?
It even might be more than the suggested 100 “remote nodes” like 200-300.
Update frequency is very low. Primary task will be Win10 to Linux deployment.
So 1 Linux image and i case this will fail, 1 Windows 10 image are store on the remote node.

I will look for documentation about both options:
1 HQ master and remote storage nodes

2 Mobile FOG deployment. If I you understand correctly, I already have that in place for testing.
A local PC with FOG server on it. Not sure how to handle the dnsmasq.
The local network has already a DHCP server (router) that I cannot configure.
I need another company for that. But ones the concept is working, I can ask to change the router options to support a bootfile and pxe-server ip.

Option 2 requires a ‘management’ solution, so maybe option 1 is in place.
We don’t want ‘boots’ on the floor and require a endpoint management solution at HQ.

[george1421]

@jaapvdpol Well lets see if we can get a few things sorted out here.

Scale: I have not seen an install yet with 100 storage nodes. Each storage node will need to be managed at some point in time.

I kind of see this as a two phase project.

1. The initial push to move from windows to linux. Will you have boots on the ground at each location to do this? If yes they can bring a mobile deployment server with them (laptop kitted with what they need for imaging). If you are trying to do this all remote I can see it taking a while to complete.
2. Post upgrade to linux. From time to time you may need to image failed computers. How will you handle that? Have a fog server at each site? Will you train local folks to be able to do this? Or is a drop ship a new system and they return the failed one for repair a bit of a better solution? Now I’ve used a raspberry pi for a fog server to image low volume systems. But that is before the prices of those critters when crazy. You think about it, with a site of 20 computers, you might, and I say ‘might’ reimage 1 computer a year. Is there value in having 100 storage nodes? I’m not saying one way or the other is right, just think about what you are trying to achieve and how much management work you are willing to do.

Now in regards to dnsmasq, that is a service we can install on the FOG server. With our configuration it will send out pxe boot information to the local subnet its connected to. You don’t need to touch anything on the router or network. So when the mobile fog server arrives, so does the pxe boot information. When the mobile fog server leaves, so does the concept of pxe booting. In this design (mobile deployment server) the target computers are not designed to pxe boot through the fog server, but instead a tech sitting in front of the computer presses F10 or F12 during booting to get to the efi boot manager where the tech picks pxe boot.

Understand I’m just trying to tell you what’s possible. Since I don’t know 100% of your use case I can only give suggestions based on how FOG works.