Issues with Boot Server - tftp timeout ITCM 12.9

[Anon Anon]

Hi,

 

we are currently facing an issue with our Scalibility-Server, that we are using for OS installations.

Boot image is Win PE 3 x64 and OS-Image is Windows 7 x64.

 

If the Client is in PXE-Boot state, a tftp timeout occurs after the DHCP information where received correctly.

 

I can retrace with wireshark, that the scalabiliy server trys to send the first block of the tftp transmission but it seems like the clients rejects them.

Atfer three retrys the tftp timeout message occurs. I also noticed that the size of the tftp block is very small (only 100 bytes), this may be the reason why the client rejects it.

 

We are at Version 12.9.0.338.

 

When I try to an OS-Installation directly on the Manager, it works like a charm. Block size of the tftp packages ist over 1000 bytes.

 

Any help is appreciated... I'm nearly desperate.

 

Thanks and Regards

 

Chris

[Matt_Godfrey]

Hi Chris,

You may want to

switch the  boot server from tftp to share access, execute the following command on the boot server:

sdbsswitch -s

 

See if that works

[Anon Anon]

Hi Matthew,

 

thanks for your reply.

I checked the mode with the -l parameter and share access is already active.

 

I think in the first part of the pxe process the transfer of the bootstrap is always a tftp connection.

 

Regards

Chris

[mccst05]

That is correct – the first part of the process, which delivers the Boot image, cannot use shares because at this point only the PXE-Enabled NIC is active and it has no support for shares, only TFTP which is part of the PXE spec.

 

There are some config parameters you may want to look at in terms of timeouts and packet sizes. They are at DSM>Scalability Server>OSIM>ManagedPC>Server, and scroll down to settings which begin with TFTP.

 

If you are seeing block sizes of only 100 bytes though it is likely the target is rejecting them as many PXE implementations do not support block sizes below 1024 bytes. This is probably being set by your network. Check the NIC settings on the Scalability Server, perhaps you can find the cause there. This is not set by the Boot Server which would have a default size of 512 unless a specific packet size is requested. It may also be due to a very poor network connection.

 

Steve McCormick, ITIL

CA Technologies

Principal Services Consultant

Tel: [phone]

[e-mail]

<mailto:[e-mail]>

[Matt_Godfrey]

Thanks for the Clarify Stephen

[Anon Anon]

Network was the right direction. We have an MTU of 1396 through the router and it seems like the pxe implementation of our clients doesn't like fragmentation.

So I used the mentioned policy to set the TFTP Max Block size.

BTW. the policy handling can really be a pain... took me hours till the Scalability Server accepted the new policy.

 

Thanks for the help.

[mccst05]

Glad I could help.

 

On the policy issue, did you make the change to the default policy? If so, you may find that you can get better results when a policy only applies to a small number of systems, in this case just the scalability servers or perhaps even a subset of them, to make a specific policy for this and apply it just to that group of machines. You can always roll the changes into the default policy later when you have a need to make a global policy change. Also, it sometimes helps to restart the policy processes on the EM/DM when you make changes to a policy. Sometimes they can go to sleep on you when they haven’t had any work to do for a long period.

 

Steve McCormick, ITIL

CA Technologies

Principal Services Consultant

Tel: [phone]

[e-mail]

<mailto:[e-mail]>

[Anon Anon]

Steve, thanks for the hint with the policy. It works now as it should.

 

Allow me to ask you one more question. The parameter TftpMaxBlksize seems only to affect the pxe phase.

We now noticed that at the start of the winpe, the script loads several files over tftp. The Blocksize there is 1456... so it often fails over our mtu 1390 network connection.

Is there a way to change the blockstize at this stage?

 

Best regards.

Chris

[mccst05]

Hi Chris,

 

There are 2 TFTP blocksize parameters: ‘TFTP blocksize limit’ and ‘TFTP specific blocksize limit’. The 2nd one, TFTP specific blocksize limit, is the parameter which is used by the ‘sdmpcimg.exe’ program which is copying those files during the imaging process. Set that parameter to match your MTU and you should be fine. Remember to make allowance for the headers (TFTP (4 bytes), UDP (8 bytes) and IP (20 bytes)).

 

Steve McCormick, ITIL

CA Technologies

Principal Services Consultant

Tel: [phone]

[e-mail]

<mailto:[e-mail]>

[Anon Anon]

Thanks again Steve. I failed to notice there are two parameters to set.

Now it works perfectly fine.

 

Our OS-Deployment ist now unstoppable

 

Best Regards

Chris

[Legacy User]

Hi.

I still have problems with this same error message If the Client is in PXE-Boot state, a tftp timeout occurs after the DHCP information where received correctly.

Any other solution?