Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
Today we installed a new fibre link down another building here is the basic layout how it stands
All green lines are fibre links
Switch C used to be served via the copper link which had to go via switch B, we wanted it to have its own connection back to the main switch A so we installed the fibre.
As soon as we plugged the new fibre in we got like a loop where switch A went mad and locked out. Could this be because the copper link was still in causing some sort of network loop? We have yet to try it with the copper link unplugged because we ran out of time, but looking at the diagram I think it should still be ok, but I may be wrong 
[Edited on 19-08-2014 by Steve]
|
oceansoul
Member
Registered: 19th Jun 06
Location: Sunbury, Surrey
User status: Offline
|
We had a similar situation at work where one of our technicians inadvertently created a loop between a number of switches. This caused all manner of network issues and crashed the network.
We have since enabled the Rapid spanning tree option in each of the switches on site to hopefully prevent this happening again. These are managed switches btw.
Although in our case the switches were already configured in a redundant ring topology (Hirschman Hiper ring). I don't know if this would make a difference in your case.
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
Is it just a case of turning on rstp on the switches?
[Edited on 19-08-2014 by Steve]
|
Dom
Member
Registered: 13th Sep 03
User status: Offline
|
What you've done is created a bridge loop (between switches A, B and C) which in-turn floods the network with broadcast packets and as you found out, switching gear will eventually shit itself and lock-up.
As Ocean mentions, enabling STP (Spanning Tree) on your managed switches should remove the redundant paths and prevent the loop.
Otherwise you can attempt to filter broadcasts (not sure if it's doable with layer 2 stuff ) or some switches support the ability to rate-limit broadcast floods - both a bit bodgy for this situation though.
Edit - the copper link between switches B and C, are you using that purely as a fall-back if one of the fibre links fails?
[Edited on 19-08-2014 by Dom]
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
yeah, will unplugging that for now help?
|
Dom
Member
Registered: 13th Sep 03
User status: Offline
|
quote:
Originally posted by Steve
yeah, will unplugging that for now help?
Yup. Although it's worth configuring that copper link properly for use as a fall-back.
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
ok cool, just want it back working reliably for now, can mess about with it out of hours at the bank holiday
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
So in terms of configuring that copper link for fall back, is it just a case of turning on STP loop preventation on the ports that copper link is connected to??
|
pow
Premium Member
Registered: 11th Sep 06
Location: Hazlemere, Buckinghamshire
User status: Offline
|
I haven't read the replied but spanning tree will sort this right out. Just remember to spanning-tree portfast (or the HP equivalent) on ports that have computers connected to them otherwise you end up with a computer not getting Group Policy changes at startup.
I assume both links are at 1GBPS? How are you terminating the fibre into the switch? SFP or a shitty media converter?
[Edited on 20-08-2014 by pow]
|
willay
Moderator
Organiser: South East, National Events
Premium Member
Registered: 10th Nov 02
Location: Roydon, Essex
User status: Offline
|
what switches are these?
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
SFP pow
Willay core is a modular HP affair not sure of the model off the top of my head that is handling all the fibre links though.
Rest of them around the school are brand new hp 1910 24g or 48g
|
pow
Premium Member
Registered: 11th Sep 06
Location: Hazlemere, Buckinghamshire
User status: Offline
|
Are they 1GBPS SFP ports or 10? If they are 1 then load balance the links, not lie redundant until someone sticks a spade through your fibre.
|
Steve
Premium Member
Registered: 30th Mar 02
Location: Worcestershire Drives: Defender
User status: Offline
|
they are running at 1gb, the modular switches can support 10 though
|
pow
Premium Member
Registered: 11th Sep 06
Location: Hazlemere, Buckinghamshire
User status: Offline
|
Load balance then 
|
willay
Moderator
Organiser: South East, National Events
Premium Member
Registered: 10th Nov 02
Location: Roydon, Essex
User status: Offline
|
its extremely easy on the HP too, but be careful as they call it trunking and in Cisco land that is something completely different.
|
pow
Premium Member
Registered: 11th Sep 06
Location: Hazlemere, Buckinghamshire
User status: Offline
|
Netgear land is even more fucked up - NTP is port aggregation!
|
