So, I bought some cheap switches and..

[Driftingashore]

Not-so-long story short: we had some very old 10/100 5 port switches sprinkled around the place and I thought I'd replace them with some cheap (Netgear) gigabit ones. I know, I know. Anyway, the first sign that things weren't fine and dandy was when the network slowed to a crawl. Suddenly network printers were showing as offline, with some printers freezing mid-job. DHCP was failing on wireless access points, and so they were rendered useless. After manually setting the speed on the core switch to full duplex gigabit for the connection, everything fixed itself.

 

Now, I'm not going to keep these switches installed, but I'm very curious about what someone with more network savvy than myself thinks may have caused this, because whatever these things were doing, they were screwing with equipment not connected directly through them - anything connected to the same core switch as that Netgear junk would be rendered useless. I've never seen anything quite so catastrophic, and honestly can't imagine what they were doing to cause such havoc.

 

Any ideas? The switch(es) are Netgear GS205, for anyone that wants to avoid them like the plague.

[MrEprise]

I'd say you bought the wrong class of Netgear switches as those look like they're meant for home use rather than business use, whereas the GS105 for example is more suited to business use and more likely to cope better on larger networks.

 

If money is tight, I'd certainly recommend sending the GS205s back and getting GS105s instead, however if budget allows I personally think you'd be better off with the 8-port version of these: HP 1810 Switch Series

[m25man]

GS205

[h=1]Technical Details[/h]

• Reliable performance
  
• Power saving technology for energy efficiency
  
• Ideal for home and small office networks
  

 

For SOHO use they will have a very limited capacity for store/forward and MAC address tables they will fill up quickly and struggle to keep track of the rest of your network.

Please don't blame the kit for something that was clearly your own mistake.

£15 for a switch, you really pushed the boat out there!

[Driftingashore]

Oh, I'm certainly aware I was pushing it a little, but 4x£15 was worth the gamble. Nevertheless, our network is small (~70 devices), and the switches were only connecting one or two devices through them. I suppose I assumed that they'd at least match devices that are approaching 10 years old.

 

Anyway, if it was simply a case of them not having the addressing space for the network, why would that have been fixed by manually setting auto-negotiate, and if it was a negotiation issue, why would that mess with unconnected devices? The reason I posted wasn't to vent about £60, more because I want to understand what may have been going on here. It seems a little weird to me, and networking clearly isn't my forte!

[FN-GM]

Please don't blame the kit for something that was clearly your own mistake.

 

No need to be so harsh...

[SYNACK]

Oh, I'm certainly aware I was pushing it a little, but 4x£15 was worth the gamble. Nevertheless, our network is small (~70 devices), and the switches were only connecting one or two devices through them. I suppose I assumed that they'd at least match devices that are approaching 10 years old.

 

Anyway, if it was simply a case of them not having the addressing space for the network, why would that have been fixed by manually setting auto-negotiate, and if it was a negotiation issue, why would that mess with unconnected devices? The reason I posted wasn't to vent about £60, more because I want to understand what may have been going on here. It seems a little weird to me, and networking clearly isn't my forte!

 

The link reset could have caused a long enough drop in the link to flush the address tables, it is more a case of how many devices are on your network total as switches keep a list of how to get to a MAC address and which port. Its not just computers but printers, other switches, wireless devices. It is possible that the address table on those was quite small and so it was exceeded and caused the lockups. If this is the case it will happen again and your only solution would be to make crazy small VLANs with the core switch to limit the amount of MAC addressed it is exposed to. That is not a good plan however, if it reoccours just take them back and get some bigger and better ones.

[plexer]

With the size of his network and the role that the devices are playing they should be fine so something is wrong somewhere.

 

Ben

[steveg]

It might be the cabling isn't up to it, which may have caused it to constantly re-negotiate the link speed. Forcing it to stay at one speed would most likely resolve this, you may drop a few packets but this wouldn't normally be noticable over a gig link. Is all the infrastructure cable cat 5e at least?

 

Steve

[steveg]

And the mac address table size on that particular switch is 4k.....

[tmcd35]

Not-so-long story short: we had some very old 10/100 5 port switches sprinkled around the place and I thought I'd replace them with some cheap (Netgear) gigabit ones. I know, I know. Anyway, the first sign that things weren't fine and dandy was when the network slowed to a crawl. Suddenly network printers were showing as offline, with some printers freezing mid-job. DHCP was failing on wireless access points, and so they were rendered useless. After manually setting the speed on the core switch to full duplex gigabit for the connection, everything fixed itself.

 

My first thought was pf a network storm. Perhaps a faulty NIC that was being stopped by the previous switch. But then I saw this line. Could it just be a simple case of the new switches having problems auto-negotiating with the core switch? You set the connection speed manually, no need to auto negotiate, everything is fine.

 

I'd monitor them for a week or two and if problems don't re surface, I'd stick with them personally.

[cpjitservices]

Netgears are 'usually' OK bits of kit. Sounds like a cable problem, how exactly have you got them linked to each other ? and do any of your links go back to the same switch if you know what I mean - this could create a switch loop. Drop your negotiation speed and see what happens... Also what switch are you using at your core ? these switches wont have spanning-tree or if they do it wont be upto much of a standard, I'd see if you have spanning tree turned on on your core switch this may eliminate a switch loop.

[Driftingashore]

No need to be so harsh...

 

Thankfully it's not my first time on a forum

 

The link reset could have caused a long enough drop in the link to flush the address tables, it is more a case of how many devices are on your network total as switches keep a list of how to get to a MAC address and which port. Its not just computers but printers, other switches, wireless devices. It is possible that the address table on those was quite small and so it was exceeded and caused the lockups. If this is the case it will happen again and your only solution would be to make crazy small VLANs with the core switch to limit the amount of MAC addressed it is exposed to. That is not a good plan however, if it reoccours just take them back and get some bigger and better ones.

 

Yeah, I'm not really happy keeping them in use, because although manual settings solve it, it only takes a small change to recreate the problem. And creating tiny little VLANs for a switch to run seems insane. Anyway, I can guarantee that we're not filling a 4k address table, or anywhere near.

 

With the size of his network and the role that the devices are playing they should be fine so something is wrong somewhere.

 

Ben

 

I happen to agree. I moved the Netgear switches onto a second switch in another part of the school with the same issue. Curiously, turning the Netgear switches on/off makes no change, but restarting the core switch made things run smoothly.. for less than an hour (although I didn't time it)

 

It might be the cabling isn't up to it, which may have caused it to constantly re-negotiate the link speed. Forcing it to stay at one speed would most likely resolve this, you may drop a few packets but this wouldn't normally be noticable over a gig link. Is all the infrastructure cable cat 5e at least?

 

Steve

 

Some of the cabling here isn't great (cat5), but as noted above, they caused the same problem on a different switch over a short 5e run. I'm not seeing any Tx/Rx errors on anything, except for a few Rx errors on a single unrelated cat5 cable.

 

My first thought was pf a network storm. Perhaps a faulty NIC that was being stopped by the previous switch. But then I saw this line. Could it just be a simple case of the new switches having problems auto-negotiating with the core switch? You set the connection speed manually, no need to auto negotiate, everything is fine.

 

I'd monitor them for a week or two and if problems don't re surface, I'd stick with them personally.

 

I'm wary of keeping gear that doesn't play nicely by default, as in a few years, either myself or the next poor sod will change something and have to figure out what the hell just happened when everything goes kaput from a minor change.

 

Netgears are 'usually' OK bits of kit. Sounds like a cable problem, how exactly have you got them linked to each other ? and do any of your links go back to the same switch if you know what I mean - this could create a switch loop. Drop your negotiation speed and see what happens... Also what switch are you using at your core ? these switches wont have spanning-tree or if they do it wont be upto much of a standard, I'd see if you have spanning tree turned on on your core switch this may eliminate a switch loop.

 

I suspected a loop too, but the network is small enough to manually inspect, and I can't find anything. We do have spanning tree, though it's turned off so I should probably set that up anyway. The core is a D-Link DGS-1210, which does have "loopback detection" enabled and shows nothing for it. It has historically been finicky with auto-negotiation (a problem connecting with a ProCurve), but the same issue with the Netgears happened on a separate switch over better cabling..

[cpjitservices]

I'd be enabling Spanning-tree, I saw at the local council someone had a mini switch and plugged it in in there office, without permission from IT (They wanted to use there laptop aswell as PC but we didnt have wifi) and it created a switch loop, slowed the network to a halt - spanning tree was enabled and left enabled and we didnt have another problem.