Multicasting with 0.33b?
-
I am having difficulties getting 0.33b to multicast images out. When I first started working on this it would load the partclone page and just sit there. I am not sure what has changed but now when I boot up the machines for a multicast job it sits at “Please Wait…”.
I have gone through the post of: [url]http://fogproject.org/forum/threads/wiki-troubleshooting-multicast.22/[/url] and ran the commands at the top of the page for udp-sender and udp-receiver. Not sure if the commands at the bottom still work for 0.33b for sending the image and receiving the image via CLI?
When I initially ran the udp-sender command the FOG server was on one subnet and the clients were on another. I could see the clients connect and receive the data that way even with a ttl of 1. I have since tried to change the server and clients to be on the same subnet just to see if my multicast routing was setup correctly and I am getting the same result.
Has anybody done multicast with 0.33b?
-
Honestly I don’t like using the multicast, so I hadn’t tested it. I do however set up groups and manage the image through the membership page. I will deploy to all 30 machines at once but only 10 of them will image at a time and from my experiences will image faster than multicast. My ~45gb image takes about 15-20 min per machine. In the past if I multicast each machine takes upwards of 2 hours and even then sometimes fails because a machine drops out or looses connection and bombs my entire lab.
Try pushing your image out as a unicast to multiple machine and see if you can get by until someone fixes the multicast, you may find it is faster like I did
. -
Jaymes,
When you were multicasting were you doing identical hardware or different hardware? What was your link speed?I have 25 identical machines to image all connected via gigabit and a 10G uplink to network. I may end up doing unicast on these machines for now but would like to know if it is a FOG server problem or a network switch related problem.
-
I did it at one point, but not with any of the revisions within the past month. Verify that the FogMulticastManager service is running - mine would do the same thing with the partclone screen until Tom helped me out with that step. If it’s not running, do a ‘sudo service FogMulticastManager start’ - that should start the service
-
[quote=“DanielJay, post: 24780, member: 23088”]Jaymes,
When you were multicasting were you doing identical hardware or different hardware? What was your link speed?I have 25 identical machines to image all connected via gigabit and a 10G uplink to network. I may end up doing unicast on these machines for now but would like to know if it is a FOG server problem or a network switch related problem.[/quote]
I have 29 identical machines at Gig link speed and 1 that is a newer model ( I usually image it manually since it is one machine and keep it out of the group so it doesn’t cause issues with hardware when pushing the image… JUST IN CASE. Plus the drivers for the machine aren’t included in the image yet due to laziness, but I have a driver disc I use for the one machine.
-
I checked to see if the service was running by doing a: “ps -e | grep FOG” from the CLI and it showed the service was running.
I decided to try downloading the image to one machine using unicast and that has failed. I am getting:
“Storage Node ‘DefaultMember’ is open, but has recently failed for this HOST
Unable to find a suitable Storage Node for transfer!”I have since opened my /opt/fog/.fogsettings file and copied the value of password into the password field of my default storage node. I have also copied that same password into the TFTP settings on the config page.
After all of that was ran I restarted my NFS and TFTP servers. I am able to open up FileZilla and FTP to the FOG server using the fog user and that same password.Any suggestions on this issue?
-
From my experience, most multicast issues come from an issue with your switch configuration.
Caution - Don’t use multicast on a production network during business hours…
I found out the hard way on this. My network bandwidth pegged out for the duration of transfer.
Multicast doesn’t care where the machines are that are getting imaged, as every port, on every switch on your network will get hit with the same data at the same time. You’re actually better off using unicast for 25 machines. Multicast is really only helpful if you have 100’s of machines to image in one night. <—emphasis on the NIGHT!
-
BigMan99211,
Multicast should not have issues if you have the network setup correctly. It is best if you use managed network switches and can run IGMP on your network. This would prevent the flood of traffic to all network ports. -
[quote=“DanielJay, post: 24808, member: 23088”]BigMan99211,
Multicast should not have issues if you have the network setup correctly. It is best if you use managed network switches and can run IGMP on your network. This would prevent the flood of traffic to all network ports.[/quote]This can also cause issues with trying to find the tftpboot server and your undionly.kpxe, be careful and take notes, I actually use the dnsmasq service because of this.
-
[quote=“DanielJay, post: 24805, member: 23088”]I decided to try downloading the image to one machine using unicast and that has failed. I am getting:
“Storage Node ‘DefaultMember’ is open, but has recently failed for this HOST
Unable to find a suitable Storage Node for transfer!”I have since opened my /opt/fog/.fogsettings file and copied the value of password into the password field of my default storage node. I have also copied that same password into the TFTP settings on the config page.
After all of that was ran I restarted my NFS and TFTP servers. I am able to open up FileZilla and FTP to the FOG server using the fog user and that same password.Any suggestions on this issue?[/quote]
Jaymes,
Any idea what might be causing this above issue?
