I've been using DS to deploy servers for about 2 years but I have recently come across a unique requirement. I need to roll out a Windows IMage that Is approx 90GB in size. Don't ask me why it Is so large as it needs to be for various reasons. Anyways, I need to roll it out to many servers over a standard 1 GB Ethernet fully switched end-to-end link. Now, I'd imagine if I simply deploy to 10 servers at a time this will generate 900GB of traffic i.e.. 10 images of 90GB.
I see that DS supports multi-casting of images. From my understanding this could mean I could deploy the image to ten machines and only Send 90 GB of traffic i.e.. Send the image once and all other machines receive from there.
Thing Is, its not clear how you do this. The only options for this I have seen are if you boot to WinPE 2 and run rdeploy.exe manually. Is there any way to do this from DS console?
Comments
There's an option I think in the control panel
If it's turned on DS should use multicasting by default but I think all the target machines have to be in the same subnet. I've never used it with WinPE but the idea is that the first client connects and will wait until it times out or other join the session up to the limit set in the Contol Panel. If two or more join, one is elected the Multicast Master and it all kicks off.
You might be able to have more control over it using scripted imaging commands from the DS console, see the DS Manual Appendix for the rdeply command line switches.
I'd maybe use a local PXE server and image store shared as "Build"; map a G: or Z: drive to "ALTIRIS_PXE_SERVER%\build" and copy the omage from that.
All my clients are in the
All my clients are in the same subnet alright. So what you are saying is this
?
If you've got multicast turned on in the Control Panel option
I think you get them to boot to WinPE by deploying the image. If multicasting is turned on in the Control Panel then each client will wait for the number of clients specified in Control Panel to join or will wait for the timeout specified in Control Panel.
There may be an option you need to select in the imaging job but I don't think there is.
You could try and run rdeploy from the command line with the appropriate Multicast switches given in the Appendix to the DS manual.
In DS 6.x, WInPE does not
In DS 6.x, WInPE does not support multicasting, only DOS is supported.
Hope this helps.
RDeploy and Multicasting
Multicasting will work in each of the automation OS's. You have to have it enabled in the console. it works exactly how andykn101 mentioned above. The key is to have the computers all part of the same job. If you kick of a job onto each of the computer individually, they will be imaged individually. If you kick off the job on 20 computers all at once, each of the clients will negotiate who will be the master. You may have a couple of masters or more, but the stream will only go to those master computers and the masters will in turn stream to the other slaves.
I can vouch for it working both Linux and WinPE in 6.9.x
Brian Hawver
Systems Engineer
Yaskawa America, Inc.
Connect Etiquette: "Mark as Solution" those posts which resolve your problem, and give a thumbs up to useful comments, articles and downloads.
Thanks Guys, I will give it a
Thanks Guys,
I will give it a try. This could save me a lot of time and bandwidth. I'll let you know ho it goes ;)
Tried that guys and it seems
Tried that guys and it seems that the image was not multicasted. I used the job scheduling wizard to deploy the job as instructed. I'm using WinOE 2 as the Automation Environment. Where do i check in the DS tha Multi-casting is enabled?
DS
You would check in Program Options, under the RapiDeploy tab. This is where you set a range for multicasting addresses.
You would also need to check in the Configuration (Start -> Programs -> Altiris -> Deployment Solution -> Configuration). Click Options and then check under the Disk Imaging tab. The Transport tab also has a checkbox related to multicasting as well but can confirm even when the "Disable multicast support (clients must connect using TCP), multicast still seems to work in our environment.
The other thing you'll want to consider is how the job is being scheduled. Are you dropping the job on to each individual computer? If so, that would mean that they would get imaged separately instead of multicasting. The best thing I can suggest is to create a folder and place all of the computer in that folder. Then drag the folder to the job or the job to the folder to schedule it.
Hope this helps.
Brian Hawver
Systems Engineer
Yaskawa America, Inc.
Connect Etiquette: "Mark as Solution" those posts which resolve your problem, and give a thumbs up to useful comments, articles and downloads.
Hi, I forgot to mention that
Hi,
I forgot to mention that all clients are in the same subnet ;). Also, here is how I deployed the job
That should ensure that they receive the image via multicast provided I have it enabled right?
My Rapideploy settings are
The Disk Imaging settings are as follows:
They are all on the same network segment on the same L2 switch so no router config needed as they are in the same VLAN. What am I missing?
Switch
What switch hardware are you using? It is possible there could be some settings that need tweaking on your switch hardware. If you are dealing with Cisco Catalyst switches, this may explain part of the problem.
http://www.cisco.com/en/US/products/hw/switches/ps708/products_tech_note09186a008059a9df.shtml
https://www-secure.symantec.com/connect/forums/setting-multicast-switch
http://www.ehow.com/how_7247095_enable-multicast-support-cisco-switches.html
Brian Hawver
Systems Engineer
Yaskawa America, Inc.
Connect Etiquette: "Mark as Solution" those posts which resolve your problem, and give a thumbs up to useful comments, articles and downloads.
Yep, its Cisco cat 2xxx
Yep, its a Cisco 2xxx alright. Not in office at the moment so the exact model escapes me :(. From my understanding, the image should unicast to 1 machine in the group/per subnet and the agent in the automation env. will multicast it out to the rest of the clients. Perhaps my problem is that IGMP is not enabled on the L2 switch and hence the multicast packet is being absorbed at the interface and not forwarded out. I'll check tomorrow. Thanks for all the help!
Correct
I believe your understanding is correct and it is possible that the switch is blocking the multicast from the master machine.
A really simple way to test this would be to put a cheap hub/switch with a few of your test machines on it which would eliminate the Cisco switch from blocking the Master client.
Brian Hawver
Systems Engineer
Yaskawa America, Inc.
Connect Etiquette: "Mark as Solution" those posts which resolve your problem, and give a thumbs up to useful comments, articles and downloads.
Would you like to reply?
Login or Register to post your comment.