up vote 2 down vote favorite
share [g+] share [fb]

Hi recently we had wast appeared to be a bad switch, we replaced the switch with a NET-GEAR JGS524 10/100/1000Mbps Gigabit Ethernet Switch unmanaged switch. The same problem is still occurring with the new switch and at random intervals the switch will loose all network connectivity. The even stranger part is that unplugging and replugging the switch does not seem to fix it but disconnecting everything and going back to an old d-link 100mbps resolves the issue. If we wait a while and put the 1000mBPS back on it works. The only logical explanation is that something is hammering the switch(s because this is not the first one or make/model to exhibit this behavior) and crashing the hell out of it to the point where it won't easily reset.

This switch is a child to another similar switch and the main network, has anyone experienced a similar issue ??

btw I am a developer not a sysadmin, but everyone is stumped on this so I'm trying to get it resolved.

Thankyou

link|improve this question
100% accept rate
Stop buying cheap hardware. – Tom O'Connor Jun 7 '11 at 16:01
Not everybody can afford high power gear to cater for abnormal situations. – wolfgangsz Jun 7 '11 at 16:03
£200/$400 isn't that much. How much working time is lost by the failure, and how many hours of expensive people time is wasted trying to fix it. – Tom O'Connor Jun 7 '11 at 16:04
Is it only this switch on the network that is affected? Are all others ok? What kind of kit is connected to the switch? – simon Jun 7 '11 at 16:07
feedback

5 Answers

up vote 1 down vote accepted

Your suspicion is probably correct, but since this is an unmanaged switch it won't be easy to verify. In order to find the culprit(s) you would need to either

  1. Replace it (temporarily) with a managed switch, and monitor traffic flow or
  2. Put a (linux) computer acting as a bridge between this switch and the uplink switch and monitor traffic using something like iptraf or wireshark or similar. However, this will only work if the overload does actually go through the uplink.

There are simple managed switches available that will allow traffic mirroring to a dedicated port. This could be used to monitor the entire traffic using something like wireshark, and these switches are not as expensive as fully managed switches.

link|improve this answer
Basically we have the servers split out to two unmanaged switches. We only experience the problem on the one switch. I originally thought about wireshark but I figured the volume of data would just be crazy as this problem occurs in monthly intervals. Pardon my ignorance on the subject but could you explain how you would go about this on a managed switch (and what features are required of the hardware). Thankyou for the helpful suggestions ... – bumble_bee_tuna Jun 7 '11 at 16:47
You need a web interface that shows the traffic flowing through the switch. We have a lot of HP ProCurve 2900 series switches that do have this feature, but they are probably too expensive for you. You could try something like the NetGear GS724-300 or the HP ProCurve V190x series. They allow performance monitoring through a web interface. I don't have the exact specs on hand, so it's probably best if you talk to a dedicated networking kit supplier in your area. – wolfgangsz Jun 7 '11 at 17:17
You will find that as you go higher on the performance/price scale, the switches come with higher spec processors and more RAM, and are gradually more able to deal with maximum traffic through all ports. In essence, Tom O'Connor is right: there comes a point where you have to invest some money in hardware that can cope with your requirements. – wolfgangsz Jun 7 '11 at 17:19
Another thing that perplexes me here even not as an sysadmin is the fact that there are not really that many boxes on this switch. Say 13 - 14 computers, and some printers ? While some of the software they use is very data intensive, it doesn't seem like a situation where the switch would be getting taxed heavily – bumble_bee_tuna Jun 7 '11 at 20:00
There is very little point in trying to figure this out without actually taking some measurements. One wrongly configured or malfunctioning device can bring this switch down completely. – wolfgangsz Jun 7 '11 at 23:05
show 1 more comment
feedback
up vote 3 down vote

Stop buying cheap hardware. Get a decent web-managed gigabit switch, say http://h17007.www1.hp.com/us/en/products/switches/HP_V1905_Switch_Series/index.aspx

Then you'll have decent quality hardware that will be able to take a thrashing that inferior hardware probably won't be able to cope with.

If it was me in your position, I'd start by replacing the switch with something decent, because that cost is likely to be less than hours of my time trying to debug a random error.

link|improve this answer
1  
There is a certain degree of truth to what you are saying we are on the second 2-300 dollar range switch as we thought the first one was defective so by now we could have bought a higher end switch. However from my end I hate to just throw money at problems – bumble_bee_tuna Jun 7 '11 at 17:01
feedback
up vote 0 down vote

Work with the vendor. I actually had a stacked set of Netgear unmanaged switches, that were dying under high load because we had Netware and Windows servers and it couldn't deal with both protocols. It took a while to get their support to diagnose that, but when they did, they replaced them for free with better (managed) switches, since the ones we had weren't under development anymore.

link|improve this answer
feedback
up vote 0 down vote

Get something online to watch traffic. Paessler has PRTG eval that has worked well for me. Works on an XP system and is not too demanding.

Is the switch on a UPS? Could you have a power issue? Low or high voltage could hang the device. Have seen this many times. Is the switch stacked with other devices. Have seen heat send the switch into a hung state as well.

link|improve this answer
feedback
up vote 0 down vote

This might be an STP problem or auto-negotiation problem.

link|improve this answer
Probably not an STP problem on unmanaged switches. Negotiation - unlikely, because a given port would work, not work, or have performance problems, but not take down the whole switch. Plus, again, it's an unmanaged switch, so there's no configuration on the switch to do. – mfinni Jun 8 '11 at 19:07
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.