FOG 1.2.0 : Multicast : PXE clients freeze, no FOGMulticastManager service

[Jonathan Cool]

Hi.
I follow the Multicast debug guide but, no better result :

• on server : killall udp-sender

• on server : udp-sender

• server say : udp-sender 20120424

• multicast logs on server say :
  "11:34:58.608529 Using mcast address xxx.xx.xx.xxx
  11:34:58.608671 UDP sender for /opt/fog/.fogsettings at xx.xx.xx.xx on eth0
  11:34:58.608694 Broadcasting control to xx.xx.xx.255

• run debug task on client and boot PXE client

• on client : udp-receiver

• client say : udp-receiver 20120424

• client say : UDP receiver for (stdout) at “ip client adr” on eth0

• and nothing happened …

[ch3i]

@Jonathan-Cool

On the wiki : FOG uses a simple queuing system to prevent its storage servers from being overworked. If you have a single FOG storage node in FOG with a queue size of 10, then this means that if you unicast an image to 30 computers, only the first 10 computers will be imaged. The other 20 computers will be waiting “in queue” for an open slot.

@Tom Elliott could you specify if it’s the main FOG server or FOG Storage node that I have the multicast service.

[Tom Elliott]

@ch3i said:

@Jonathan-Cool

On the wiki : FOG uses a simple queuing system to prevent its storage servers from being overworked. If you have a single FOG storage node in FOG with a queue size of 10, then this means that if you unicast an image to 30 computers, only the first 10 computers will be imaged. The other 20 computers will be waiting “in queue” for an open slot.

@Tom-Elliott could you specify if it’s the main FOG server or FOG Storage node that I have the multicast service.

Technically, all systems have the Multicast Service on them, it’s only the Master Storage node in the group that actually generates the Multicast command line informatoin.

[ch3i]

@Tom-Elliott If a NAS have not FOG installed as FOG STORAGE he can’t multicast if is it MASTER STORAGE ?

[Tom Elliott]

@ch3i said:

@Tom-Elliott If a NAS have not FOG installed as FOG STORAGE he can’t multicast if is it MASTER STORAGE ?

That is about correct. Unless the NAS box is of a distribution that natively can support a real fog installation or somebody is able to port the udp-cast information and the service files to the NAS box, most likely it will not be easily possible.

The best solution, in this case, is to simply make the NAS a “mounted” filesystem on a FOG System that is capable with a proper linux/(maybe bsd) distro.

[Jonathan Cool]

Hi !
Thank you for your answers Tom and ch3i.
If i understood, the only solution is to delcare the NAS directly on the FOG server by using the /etc/fstab ?
Like this : NAS_ip:/volume1/images /images nfs defaults 0 0
and use the DefaultMember storage ?

[Tom Elliott]

@Jonathan-Cool That is correct. Though there is a potential issue. As the NAS is shared via NFS to the FOG Server, then the FOG Server is trying to share it’s NFS to the other systems. This may be a problem because NFS can’t be shared if it’s mount point is already NFS.

[Jonathan Cool]

@Tom : Yes. I know this issue and it can be solved by using unfs3 … but, i think, with unfs3, the average transfer will be slow.
Now, i’m not sure if multicast is necessary …
If i have time, i will compare the average transfer time between “unicast” and “multicast unfs3” and will post results here.

[Tom Elliott]

@Jonathan-cool I think a better option, maybe, is to mount the NAS as SAMBA/CIFS and then you would not have much hit on network performance. To add to that, if you’d be able to upgrade to Trunk, you’ll find the init’s perform Way better. I know it’s scary to jump onto a dev release, but it works.

[Jonathan Cool]

@Tom Elliott : Many thanks for your answer.
Now, i will use my brain and think if multicast is really necessary …
In french, a proveb say : “Le mieux est l’ennemi du bien”.

In few days, i will discuss with my collegue of the situation and we will choose a good decision about that.

PS : Learn guitar is hard but it’s a very cool hobbie