Issue with Cisco Switch, 2950 VS 3560
We upgraded fog from 0.32 to 1.1 and now we have a big issue. In our network we have couple switch model and for a unknown reason it is not working anymore on the older one.
Before the upgrade, version 0.32, i was able to boot PXE with every kind of PC on every switch, cisco 2950, 2960, 3560 and 3750. Now, after the upgrade, for the old PC i am able to boot PXE from any switch but if i try a more recente pc, for exemple a laptop Lenovo L520, it is working on the cisco 3560 and 3750 but it is not working anymore on the 2950 and 2960 … PXE boot is impossible…
Can someone tell me what changed in this upgrade that cause the issue ?? Is there a solution easy to apply ?
Thank you !
Shelly12 last edited by
The Cisco Nexus 6000 range contains two models, the 6001 model and the 6004 model.They can be used as layer2 and layer3 switches and can aggregate traffic from the Fabric Extenders (FEX) for different blade-server systems. Both models support either front to back or back to front airflow and they do support Fibre Channel over Ethernet in combination with a ‘full’ FCoE switch.
[URL=‘http://vube.com/INDIGO MULTIMEDIA/cisco-642-813-certification/B9ZhkmAAmC’]vube.com/INDIGO MULTIMEDIA/cisco-642-813-certification/B9ZhkmAAmC[/URL]
[quote=“csrs, post: 30661, member: 24376”]Thank you all for the reply, as i can see for now, 99% of the issue was because of portfast … I am working in educational organisation and we have over 50 schools to support from a centralized datacenter connected with black fiber over each buildings. In lower grade school we have alot of hubswitch and it is not possible to set the switch port to portfast … In the 0.32 version of fog everything was working well … i am thinking about a downgrade but in the same time i assume the next kernel will not be compatible for futur upgrade …[/quote]
This is odd, I was able to use the new fog server in the same setting as my 0.32 server. But I also use DNSMasq to boot my machines. You can look into DNSMASQ, it’s a quick easy set up and if it doesn’t work it’s just as easy to get rid of.
Thank you all for the reply, as i can see for now, 99% of the issue was because of portfast … I am working in educational organisation and we have over 50 schools to support from a centralized datacenter connected with black fiber over each buildings. In lower grade school we have alot of hubswitch and it is not possible to set the switch port to portfast … In the 0.32 version of fog everything was working well … i am thinking about a downgrade but in the same time i assume the next kernel will not be compatible for futur upgrade …
mark graham last edited by
on access ports, the best thing to do is to use the command “switchport host” on each interface thats an access port, this will for sure enable portfast, change the port to access, and clear any port channeling. the reason this happens is the switch has to make sure its not a switch or a device that could cause a loop by sending BPDU’s. this is mostly used for the election process for spanning tree.
on some of the older switches and or older IOS images, this may or may not be in as a feature. at that point you would have to use spanning-tree portfast instead.
It is also now in the wiki under the How-to-guides–>[url]http://www.fogproject.org/wiki/index.php/Not_passing_PXE,_or_ProxyDHCP...NO_PROBLEM_Cisco[/url]
Did you use the search feature?
Here is a troubleshooting post I made and even left my changes so others could be successful in setting up their networks. Please note, I am NOT a cisco engineer, and I am by no means claiming to be an expert, but if you want some steps to try:
Thank you for the reply Tom, I will try it and let you know if its work or not. We are using STP but probably with different setting on the 3560 who are working against the 2950 who isn’t.
I will do couple test and put the result here.
Do you guys use Spanning Tree Protocol?
iPXE does not support STP so well, as it has to recall for DHCP. During this, the STP interferes by disabling the port, reforwarding to get dhcp, then allow traffic to pass.
Try enabling portfast or disabling STP altogether and I imagine you’ll see higher success.