Configuring FOG - Storage Nodes and Multicast

[MCSDFog]

Hello,

I found in my environment, it is helpful for multicasting if I have a FOG server on each of our vlans (two we use for imaging). However, this approach introduces the problem of separate databases; separate host registration, images etc.

It is my understanding a possible solution/better approach would be to have one FOG Server (normal install) and multiple FOG Storage Nodes. My goal for this approach would be to have one FOG server (normal install) and a FOG Storage Node for two of our vlans (vlan 1 and vlan 2).

However, I am a bit unclear about multicasting and Storage Nodes. How can I make sure that computers on vlan 1 connect to the FOG Storage Node on vlan 1 and computers on vlan 2 connect to the FOG Storage Node on vlan 2 when multicasting?

Thanks

Randy

[george1421]

I guess first of all, you can only multicast from a master FOG server. The storage nodes don’t have multicasting capabilities. So storage nodes are limited to unicast imaging.

Secondly multicasts are a bit different that unicast messaging. Typically routers will block a multicast message from crossing the network subnets. To this you need a router that supports multicast messaging or a multicast router (mrouter), or a multicast proxy agent running on your vlan router (its akin to the dhcp-relay/dhcp-helper service you are using for dhcp messaging). If your current infrastructure doesn’t have a mrouter then you “might” spin up a pfsense router and only use the multicast proxy service to forward the multicast messages between the vlans.

As for your network infrastructure, you probably already have igmp snooping turned on for the vlans that have multicast traffic. This will turn on sparse mode instead of dense mode multicasting.

Edit: Well I’ll be… I did have a tutorial on this with pfsense: https://forums.fogproject.org/topic/10117/fog-casting-across-vlans-subnets

[MCSDFog]

Hey george1421,

Thanks for the response.

To clear things up a bit. We do have multicast working across vlans, the issue is more the reliability. On one vlan only certain models (Dell e5430s, Dell 3380s) will multicast with others (Dell 3340s, Dell 3350s, Dell 3390 2in1) get to the gparted screen and just sit like they are waiting additional computers to join. Then on both of the vlans if you need to cancel the multicast tasks and try to kick if off again, it seems to be hit or miss if the host get past the gparted screen and start imaging. Seems I need to wait a certain amount of time before trying to kick off the job again. I haven’t nailed down that minimum wait time…seems to be 15 min or less. If I do it too soon it just sits at the gparted screen.

So that is why I decided to look at the Storage Node approach. I then came across the wiki article about Storage management.

https://wiki.fogproject.org/wiki/index.php?title=Managing_FOG#Storage_Manangement

In the Storage Management section it talks about Storage Groups and multicasting via the Storage Group’s Group Master Node. I was unclear if I could make sure hosts on a vlan hit the Group Master Node on that vlan. Maybe I was adding wishful thinking.

Thanks

Randy