Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
Networking (Apple)

Journal Saint Aardvark's Journal: Network problems 15

Journal by Saint Aardvark
About three times in as many weeks, we've had a switch at work seize up. We've got three Etherfast 4124 switches from Linksys cascaded together, and something is causing them to just freeze. It's only ever one -- whatever's causing it doesn't traverse the link between two of 'em. Symptoms: lights freeze up, ping times suddenly shoot right up and then all traffic on the switch stops altogether. If you find the port that's causing problems quickly enough, you can yank out the cable when ping times shoot up, and everything's fine; otherwise, you have to power cycle the switch.

Twice, I've been able to track it down to one network plug in particular that's causing problems. It's not the same port on the same switch every time, nor even the same switch. The only thing that has been common between the two has been a cheap SMC 8-port switch at the other end (not the same one each time). Because the 4124 is a dumb switch, I can't do much troubleshooting on it. And I'm unable to duplicate the problem -- once the problem port is disconnected for long enough, things settle down and reconnecting it doesn't cause problems. No chance to run tcpdump, or otherwise figure out what the hell is going on. I've replaced the SMC switch and I'm going to be calling DLink to see what they think.

As a result, I've been looking at managed switches over the last few days, and holy crap are they cool. I knew that a managed switch was The Right Thing to get if you were serious about a network, but it wasn't until I started reading the manual for one of them (the Tigerstack 3 from SMC; despite the possible problems with the cheap switch, this one seems to have pretty good reviews) that I realized just how much you could do: port mirroring (which'd help if one port is causing problems), VLANs (keep our Windows traffic separate from our Unix traffic), SNMP monitoring (watch for sudden spikes), blah blah blah. I'm preparing to tell my boss that we should really spend some $ on upgrading.

So some questions for anyone that can answer:

Has anyone had a problem like this before? Is this just cheap equipment? What kind of traffic could freeze a whole switch? Since the problem keeps going away once I disconnect the problem port, and since I really can't keep bringing down the network while I figure out what's going on, how can I track things down?

And does anyone have any recommendations on managed switches? Like I said, the Tigerstack seems pretty well-recommended, and I think it's within our budget. What do you think?

As you can probably tell, I am learning this as I go along. I've made one mistake already; I'd like to avoid a second.

This discussion has been archived. No new comments can be posted.

Network problems More

Network problems

Comments Filter:
  • Cisco Catalyst switches are the bomb.

    (yeah, I'm a shill for Cisco stuff)

    Can handle many vlans, you can do port level debugging (assuming you can console in) which is really nice to see what exactly is going on wrt a specific port.

    Need a little more info on these specific ports. Are they pcs, servers, other switches? What else is connected. Can you do a port level display of the in/out packets or errors?
    • Hiya, Em -- I was hoping I'd hear from you.

      Cisco Catalyst switches are the bomb. (yeah, I'm a shill for Cisco stuff) Can handle many vlans, you can do port level debugging (assuming you can console in) which is really nice to see what exactly is going on wrt a specific port.

      I figured you'd recommend Cisco. :-) The price though is quite a jump; 700 Canuckistan pesos for the SMC vs 2k for a 24-port Catalyst. Ouch. I'm hoping that I can get away w/less. (I'll almost certainly have to anyway.)

      Need a

      • Hmmm.

        That is odd. There are a couple things I'd look at, and plz forgive me for stating what will probably be obvious ;)

        1) check the windows box for viruses. There are a couple out there now that spew "broken" pings in a broadcast and it absolutely crushes down switches and routers.

        2) check the firmware settings on the switch, see if the provider has listed any bugs or anything...this is a stretch since you said they weren't manageable.

        3) download a freeware sniffer and see if you can see anything unu
        • check the windows box for viruses. There are a couple out there now that spew "broken" pings in a broadcast and it absolutely crushes down switches and routers.

          Huh, didn't know that. Checking now. Never omit the obvious with me. :-)

          2) check the firmware settings on the switch, see if the provider has listed any bugs or anything...this is a stretch since you said they weren't manageable.

          Just called Linksys and there's no way to upgrade the firmware.

          3) download a freeware sniffer and see if you c

  • I second the vote for cisco. I've used both cisco and netgear managed switches, the netgear only allows you to mirror one port to another, the cisco allows you to mirror all ports to one. Also, snmp on the netgear is VERY unreliable, using mrtg the netgear wouldn't respond to snmp queries about 75% of the time, never had a problem with the ciscos. Of course the price difference is about $1000.
  • Yes, I've had problems like that before. What are all the lights doing when the problem is occuring? I wonder if you aren't getting a loop on occasion.

    BTW, you say "cascaded". Are you familiar with the rules for chaining bridges and repeaters?
    • "BTW, you say "cascaded". Are you familiar with the rules for chaining bridges and repeaters?"

      Does that apply to switches? I thought that applied only to hubs and other "dumb" devices, correct me if I'm wrong.
      • Nope, it applies to repeaters and bridges, smart or dumb. Note that I'm using more precise words here, as people sometimes aren't specific enough when using the words "hub" or "switch".

        You get 4 repeaters in a row, and 7 bridges.

        If the three devices are full per-port switches, then there should be no problem chaining them (though it may not be ideal for performance.) If they are repeaters (say for example, dual-speed hubs), then a repeater hanging off both ends of the chain is going to be 5 in a row.
    • Yes, I've had problems like that before. What are all the lights doing when the problem is occuring? I wonder if you aren't getting a loop on occasion.

      The flashing slows down, then stops altogether and lights (for ports plugged in) remain solid.

      BTW, you say "cascaded". Are you familiar with the rules for chaining bridges and repeaters?

      Mm, I think I'm familiar...no loops (no more than one path from A to B) and no cables > 100m. Is that the sort of thing you're talking about?

      FTR, the three switche

      • It could be a hardware failure, and no predicting wha could happen then. However, solid lights are also what you get when you just fill the pipe as full as possible.

        Is it just the light on the bad port, or are other ports having the same problem?

        All computers are having problems when that happens, or just the one(s) in the flaky port?

        The rules are 4 repeaters and 7 bridges (plus, as you asked, things like cable length, but I wasn't asking specifically about that.) There are also rules about how many pa
        • It could be a hardware failure, and no predicting wha could happen then. However, solid lights are also what you get when you just fill the pipe as full as possible.

          If it's a hardware problem, all three switches are having it; I was able to crash/freeze all three by moving the problematic cable around. Hadn't thought of traffic levels, but I was just trying to get things working again and didn't think to check for that. If/when it happens again, I'm going to be finding an affected computer and running

          • It does sound a bit like you've got someone stuck sending at maximum speed, on occasion. Especially if the problem is following a particular wire around.

Slashdot Top Deals

Real Programmers think better when playing Adventure or Rogue.

Close