Would not agree on the portfast thing -- if you are running "Rapid
spanning tree" (also known as RSTP/802.1w, and you should be) then
portfast is a good thing (declares the server ports as "edge ports" and
turns off the typical 802.1D STP "blocking > listening > learning >
forwarding" delay, which can cause up to 50sec of no network
connectivity when the nodes first boot.) We have had problems on Dell
switches with PXE booting nodes where the nodes have bonded links to the
switch -- turning portfast on all link connections solved the problem
for us.
Also, these are fairly old Dell switch platforms; are they running
latest firmware? Dell has had a lot of patches to their switch OS's, and
it pays IMHO to be on the latest revision.
Here's how to check which version of spanning-tree you are running
(example taken from a Dell PC5548 we have here):
sw# sh spann
Spanning tree enabled mode RSTP <<< mode should be "RSTP"
Default port cost method: long
Root ID Priority 24745
Address 44:d3:ca:7e:3c:c0
Cost 200019
Port gi1/0/48 <<< this port should be the
uplink port to your core
Hello Time 2 sec Max Age 20 sec Forward Delay 15 sec
Bridge ID Priority 32768
Address 5c:26:0a:a9:04:2b
Hello Time 2 sec Max Age 20 sec Forward Delay 15 sec
Number of topology changes 3893 last change occurred 48:03:39 ago
Times: hold 1, topology change 35, notification 2
hello 2, max age 20, forward delay 15
Interfaces
Name State Prio.Nbr Cost Sts Role PortFast Type
--------- -------- -------- -------- ------ ---- --------
-----------------
gi1/0/1 enabled 128.1 20000 Frw Desg Yes P2P (RSTP)
<<< "Portfast" state should be (IMHO) "Yes"
gi1/0/2 enabled 128.2 20000 Frw Desg Yes P2P (RSTP)
gi1/0/3 enabled 128.3 20000 Frw Desg Yes P2P (RSTP)
gi1/0/4 enabled 128.4 20000 Frw Desg Yes P2P (RSTP)
(...etc...)
If you need to change the switchports to use portfast, here how to
change them en-masse:
sw# configure
sw(config)# int range gigabitethernet1/0/1 - 47
sw(config-if-range)# spanning-tree portfast
sw(config-if-range)# spanning-tree bpduguard enable
sw(config-if-range)# end
sw# copy run start
Overwrite file [startup-config] ?[Yes/press any key for
no]....05-Dec-2006 01:50:36 %COPY-I-FILECPY: Files Copy - source URL
running-config destination URL flash://startup-config
05-Dec-2006 01:50:45 %COPY-N-TRAP: The copy operation was completed
successfully
Copy succeeded
I also like to use another param on switchports running portfast named
"BPDU Guard" -- since portfast pretty much turns off spanning tree on
the port, you may be susceptible to bridging loops if a redundant
connection is cabled up between switches on the portfast ports (usually
STP would detect and mitigate this by shutting down one of the redundant
ports.) When a switch port comes up (and spanning tree is enabled, which
it almost always should be unless you have a VERY good reason not to),
it sends out a "bridge protocol data unit" frame which is used to
calculate the STP topology; if a port with the "spanning-tree bpduguard
enable" configuration receives a BPDU, then it will shut down the switch
port as a precaution, since STP is disabled for this port, and now it
detects that another switch has been connected to it.
Anyways, hope this helps.
Best,
Will
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of John Stoffel
Sent: Friday, January 11, 2013 3:42 PM
To: Morse, Richard E.MGH
Cc: [email protected]
Subject: Re: [lopsa-discuss] Network configuration question
Richard> Hi! I have a problem that I was hoping I could get some advice
Richard> on. In particular, where to start looking...
Richard> I inherited a Rocks Cluster which I am trying to make
Richard> semi-usable. There are 31 nodes in the cluster, so in order to
Richard> create the cluster network, two Dell switches are used -- a
Richard> PowerConnect 5224 and a PowerConnect 5324. The head node and
Richard> most of the computer nodes are plugged into the 5224, and the
Richard> rest are plugged into the 5324 which then is plugged into the
Richard> 5224. The problem that I'm having is that the clients on the
Richard> 5324 cannot be PXE booted properly. If I plug them into the
Richard> 5224, let them boot and install, and then plug them back into
Richard> the 5324, they work correctly, until the next time I need them
Richard> to reboot (to add a package, or something).
Richard> I assume there is some kind of packet that is not being
Richard> properly forwarded by the 5324 (or not being allowed by the
Richard> 5224). What I would like is to convince the two switches that
Richard> they are really one big(ger) switch, so they don't block
Richard> anything from going around. Unfortunately, I really don't know
Richard> much about networking beyond the basics -- in particular, I
Richard> don't know the terminology. Looking at the configurations of
Richard> the switches, I see things like "spanning-tree", "bridging",
Richard> "trunks", etc.
Richard> What should I be looking at? I tried to Google various things,
Richard> but I'm not convinced I even know where to start -- at least,
Richard> nothing I've found seems to address my question, or at least
Richard> make sense to me.
Try checking that port-fast is/isn't turn on for the ports. We have a
problem here with our DELL 5224s not working right with our Cisco 4507
running IOS showing the same symtoms. So far I haven't beaten on it
enough to chase down the real config issue, mostly because I only
kickstart once in a while and I don't run the switch and don't want to
run the switch.
Nate also has some good thoughts here, and more details too. Good luck.
John
_______________________________________________
Discuss mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
http://lopsa.org/
_______________________________________________
Discuss mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/discuss
This list provided by the League of Professional System Administrators
http://lopsa.org/
