[K12OSN] Setting up a Managed Switch

["Terrell Prudé Jr."]

Carl Keil wrote:
>>
>>
>> Asmo Koskinen kirjoitti:
>>  
>>
>>>> I do use Ubuntu 6.06, but solution should be the same - add a line
>>>> to file:
>>>>
>>>> /tftpboot/lts/[--]/pxelinux.cfg/default
>>>>
>>>   
>>
>> Mine looks like this (LTSP 4.2):
>>
>> prompt 0
>> label linux
>>  kernel bzImage-2.6.16.1-ltsp-2
>>  append rw root=/dev/ram0 initrd=initramfs.gz
>> MOPTS=nolock,ro,wsize=2048,rsize=2048
>>
>> Best regards Asmo Koskinen.
>>
>
>
> Thanks so much for the replies.  (And on the weekend, no less)
> The clients boot up now, but it takes almost 10 times longer than
> before on the other switch.  Performance seems fine once they've
> booted.  There are a lot of messages throughout the bootup process
> saying "nfs server not responding" followed by "nfs server OK".  It
> eventually gets there, but I can't really function with bootup times
> this long.  One thing I noticed is that it said that it had only
> negotiated 100BaseT, half duplex.  Could that be causing the
> slowdown?  I told the switch to auto negotiate, but to default to full
> duplex 100, should I force it to full 100?
>
> Any further tweaks I can try to speed up the boot times?
>
> Thanks,
>
> ck 

Nope, don't force it to 100Mb/FDX if you can at all avoid it.  The IEEE
rules for autonegotiation state that if one side autonegotiates, but the
other side isn't, then the autonegotiating side MUST fall back to half
duplex, at the speed of the locked-down (i. e. non-negotiating) device. 
However, if both sides are capable of autonegotiating to FDX, then
that's what MUST happen.  I've got Cisco Catalyst, Bay/Nortel, and
Amer.com switches that do this perfectly, every time, no fuss.  My NIC's
are 3Com 3C905 series, Intel Pro/100 series, and Realtek 8129.

I'd be getting on the phone with 3Com to troubleshoot this, especially
if things were working right with your previous switch.

--TP