Can't Netboot more than 11 computers at a time

[catfeetstop]

I'd like to be able to Netboot around 50 machines at a time but for some reason we can't boot more than 11 at a time. Sometimes we boot a few more than 11 but never more than 13. 11 seems to be the norm. When we attempt to boot more than 11 it looks like their going to work but then they eventually just reboot to the internal HD. Any ideas?

- jamie

[catfeetstop]

I think it has something to do with the "diskless" Netboot option.

[Chris]

Just an idea, have you checked how much free disk space you have on your NetBoot server when these 11 clients are booted?

[catfeetstop]

I have lots of space available still. Though, I have intermittently gotten a Startup Disk is full message. What are you thinking?

[jhalvorson]

Another item to check is the number of available IP addresses in your subnet. Might need more IP's or a shorter lease time.

[jhalvorson]

To address the startup disk full message, you can pad the netboot image. Below is an example.

1. Open up a terminal window.
2. Type:

   sudo hdiutil resize -size 13g /Library/NetBoot/NetBootSP0/MacNetBoot10.7.nbi/System.dmg

Check the size of your dmg first so you know how much to increase it.
The netboot images I capture for 10.7 + Casper Imaging are around 11GB using System Image Utility with 10.7.3. I don't completely understand the logic on picking a number, but if I pad it to 13GB it works for us. Setting it to 12GB didn't increase the available space enough.

[catfeetstop]

Thanks for the suggestions. I increased the Netboot image size but I can still only boot 11 machines. The Netboot image lives on one of the secondary hard drives on the server. Next, I'm going to try moving the .nbi to the system drive. I'll report back.

[catfeetstop]

Nope, I can still only boot 11 machines. Any more suggestions? I'm sure it's something really simple and stupid, it always is.

[acdesigntech]

What speed is your network? 50 clients requires 100MB switched ethernet at minimum... 10 MB non-switched maxes at 10 clients, but I find these are fuzzy numbers so the variation you're seeing might be due to available bandwidth and network saturation at any one particular time. In fact, I don't have much luck with over 25 clients on 100MB switched. It should be more along the lines of 1000MB switched for 50 clients. duplex also matters, if your network speed is throttled at half-duplex, that's going to hamper the number of clients you can boot at any one time. Keep in mine that this isn't just client network connection. If you have a bottleneck between your server and the clients, you'll see this as well.

Padding the netboot image won't have an effect on how many clients you can boot, BUT the diskless option allows NB clients to create shadow files for required read/write operations... It's essentially you have a large amount of free space available on your server for this purpose.

The netboot server does not delete these files when a client is no longer netbooted. These files live in the /Library/NetBoot/NetBootClients0 directory (or wherever you have your NetBoot folder redirected). I have 71 client shadow image files in there taking up a little over 10 GB right now... Most of the time these are zero unless a NB client is using them, but sometimes they have a non-zero size.

It's safe to go through and delete them as the client will just recreate if needed on reboot to the NBI. The same mac address gets the same client name every time, so if you do not delete the file, it will be reused by the same computer.

New clients will create new shadow files, so this can grow very quickly.

Also, (and I'm sure you're aware of this) netboot relies on bsdp (bootp) that is a broadcast protocol that won't hop subnets by default unless you're routers/switches/what have you are specifically configured to do this. If you have some sort of switching/routing funkiness going on whereby some clients are put into a different subnet than the server, you can try to netboot and the client will make an attempt, but fallback to the internal HD.

And finally, to quote jhalvorson,

Another item to check is the number of available IP addresses in your subnet. Might need more IP's or a shorter lease time.

Check your DHCP leases. I had the exact same problem on my staging subnet where leases for some reason were set to 8 days. Some days 20 clients would boot, some days only 8, some 30. Made for a lot of fun during imaging time ;)

[tlarkin]

Definitely check out your DHCP settings. Lease time, and how many IP's you lease out. I have seen this at my old job when we were imaging Macbook Airs off of multiple Mac Mini servers. Netboot requires DHCP to be handed out by something, and we were using cheap consumer routers (for NAT) and a smart gig switch to keep imaging off our network for our project. They only lease like 50 IPs and the lease time is 24 hours. So once you give out all 50, nothing could netboot. We changed the lease time to 10 minutes and set the pool to 150, and our problems went away.

[catfeetstop]

Alright, thanks for the suggestions. I'll talk to my network manager to see if we can dig into things. I'll let you know.

[jshipman]

Check your AFP Settings on your server. It's under the server admin app and then if you select the server its in Access. You want to make sure that ll users are selected under the AFP settings. This little setting was a problem for me...

[tlarkin]

Oh also, another caveat we ran into was storm control. When we had our dinky routers out in the buildings only there for NAT all requests to the JSS for auto run data went through the same network port. Storm control did not like this, and shut down the port thinking it was a storm or a loop that was causing all that traffic through the same port. We had to have our network admin turn off storm control on the ports we were using to have our clients hit the JSS for auto run data.

Let us know if you still cannot get it fixed.

[Chris_Hafner]

FYI The max number of AFP connections in Lion server is located in the NetBoot General Settings tab just in case you're running 10.7 server.

[technicholas]

I have the same problem! HELP

We have two different machines. One server is for netboot and one houses the image for casper. We tried again today with a brand new Mac Mini i7 8GB of ram for the netboot and on a 1GB port and we still have problems The Casper server is a Mac Pro with 16GB of ram so I don't think that's a problem. The Casper server is bonded ethernet at 1GB. We started the machines and netbooted of them and 4 out of 30 got into the netboot we get past the flashing globe then get into the spinning globe by the apple then it sits there for a while and reboots back into the factory image.

Once it gets into the netboot it launches the JAMF software and starts imaging from the Casper server.

I don't think its a DHCP problem we have enough addresses. They are all Macbook Pro 13 in with Mac OS X Lion

HELP!

[simbimbo]

I just ran into a similar problem, it turned out the be Spanning Tree on the switch ports. The switch wasn't coming up quick enough to allow the NetBoot to succeed. Just a thought..

[fabian_ulmrich]

Are there any updates on that post? I never had that issue coming up until I upgraded to Maverick Server and the Server.app v3.02. Since the upgrade I experience the same problem (not more than 11 clients at the same time), and sometimes a weird other issue pops up which apparently is related to the Shadow file location, although I am booting in diskless mode. Any suggestions if this post is still alive?

Thanks & Cheers

[jwojda]

are these 2010 models?

[fabian_ulmrich]

@jwojda So I am having the latest MacBook Airs (Identifier: MacBookAir6,2) and my NetBoot Server is a MacMini Server!

[charliwest]

I think the solution you are looking for is here

http://www.macos.utah.edu/documentation/administration/setup_netboot_service_on_mac_os_x_10.6.x_client.x_client/setup_netboot_service_on_mac_os_x_10.6.x_client-diskless_netboot.html

Its due to AFP limitations, this boots the disk into RAM and stops using the netboot server after that

[fabian_ulmrich]

@dwest Yeah, that article is recommended by JAMF support as well. I did this to my rc.netboot, but it's not changing anything. Sometimes it's even not booting from that Image. It drives me crazy and I never had those issues before. Need to image 130 MBA until next week :(

[fabian_ulmrich]

Ok...one step back. My fault, forgot a ` in one of the commands. So booting works fine without any weird issues. But still just 11 machines at the same time :/

[jhalvorson]

I had been replacing the rc.netboot file for prior versions of OS X and with 10.9 it didn't seem to work right. Then I found post on jamfnation that pointed out that two lines of the rc.netboot script need to be changed to support 10.8 and 10.9. I am not able to get that info or link to you now, but it's in JAMFNation. I'll follow up to this post later this weekend.

[Kumarasinghe]

May be @jhalvorson is referring to this;
https://jamfnation.jamfsoftware.com/discussion.html?id=6443#responseChild33353

[Chris_Hafner]

I don't know if this is worth pointing out, but I've been successfully netbooting 30-50 machines for years from Apple NetBoot servers... over NFS. I know this isn't helping but I keep seeing this AFP reference to a limitation of 10 netboot shadow file volumes being hosted on the OSX Server because of an AFP limitation. Perhaps this is a configuration I've never used for some reason? 10.8 and 10.9 NetInstall only allows for nbi's to be hsoted via HTTP or NFS anyways. What am I missing?

I will also admit freely that, in the past, I've never had to modify the rc.netboot file (until recently), nor have I had to expand the netboot volume or resort to RAM Disk... though I've played with the performance differences. So while I'm throwing myself out here, what sort of performance increases do you folks experience in large scale netboot environments by:

1) Expanding the netboot image? (FYI, I DO NOT create a minimal NetBoot by stripping away unneeded files nor do I use Composer to compress the image.)
2) Utilizing RAM Disk for the shadow volumes. 3) Using a minimized .nbi?

[wabit]

I've experienced the same issue, net boot can't serve more than 8 clients, try different net boot images on different server, same result. do you have any final workaround about this issue?Thanks-Kinston

[catfeetstop · Accepted solution]

@wabit I saw you made another post and they replied with pretty much what I was going to say. The modified rc.netboot made with AutoCasperNBI helps. I also think part of the issue was with our network speeds. I originally posted this when I was at my old job. The network there was much slower. It took about 30 mins or more to image a single computer. At my new job it only takes about 5 mins to image. I think the network speed probably had something to do with it too.

I'm going to mark this post as resolved. Hopefully you can get it straightened out in your other post. Best of luck to you!