Linux - NetworkingThis forum is for any issue related to networks or networking.
Routing, network cards, OSI, etc. Anything is fair game.
Notices
Welcome to LinuxQuestions.org, a friendly and active Linux Community.
You are currently viewing LQ as a guest. By joining our community you will have the ability to post topics, receive our newsletter, use the advanced search, subscribe to threads and access many other special features. Registration is quick, simple and absolutely free. Join our community today!
Note that registered members see fewer ads, and ContentLink is completely disabled once you log in.
If you have any problems with the registration process or your account login, please contact us. If you need to reset your password, click here.
Having a problem logging in? Please visit this page to clear all LQ-related cookies.
Get a virtual cloud desktop with the Linux distro that you want in less than five minutes with Shells! With over 10 pre-installed distros to choose from, the worry-free installation life is here! Whether you are a digital nomad or just looking for flexibility, Shells can put your Linux machine on the device that you want to use.
Exclusive for LQ members, get up to 45% off per month. Click here for more info.
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Rep:
Gigabit switch causing network to freeze.
We have an old Zonet 10/100 switch in our home network that locks up whenever there's a power glitch longer than about a microsecond. (OK. It's not that bad but if the lights flicker, I usually end up going down to the basement to see if the switch is hung.) None of the computers or peripherals are affected by the glitches, BTW. Power-cycle the switch and everything starts working again. After putting up with this for months I finally picked up a Netgear 8-port 10/100/1000 switch to replace the Zonet. Unfortunately, when the Netgear switch is in the network, everything slows to a crawl before completely freezing up. Pinging other systems barely works. Pinging by hostname would not work. Pinging another system by its IP address worked a little but packet losses of 70%-80% were the best I saw. At first. Then it gets to where the loss is 100%.
All systems were rebooted after the Netgear switch was inserted into the network. The systems that were capable of gigabit connections autosensed the new switch and set their port speed accordingly. And that's about all that works when the Netgear switch is present. (The Zonet is back in the network now.)
One thing I have not tried is forcing the port speed on the gigabit-capable systems to 100Mb and re-inserting the Netgear switch into the network to see if the problem if due to running some of the ports at 1000Mb.
The systems on the network are running various versions of Linux (with different kernel versions, of course). Most are due for upgrades to get them current but I was planning on hitting the systems that had gigabit-capable ethernet interfaces thinking that a.) 100Mb speeds have been around forever and no problems have been seen when everything was running at 100Mb in the past and b.) the gigabit support for the older kernels should be better (one system only supports an MTA of 1500 max at 1000Mb).
Any ideas as to what could be causing the network freezes?
Could the gigabit support for the older kernels be poor and cause these network freezes?
Do Netgear switches have any history of freezing, especially when systems are using different network speeds?
Depending on the age of the switch(not dependent on when you bought it) Netgear was shipping the same power supply on the 8 port switch as on the 5 port switch. Two or three years ago(maybe longer) this was a real and pretty common problem. Usually the excess power draw(amps) drops the voltage so far the unit just stops working, but sometimes the unit(not the wall wart) just got hot and burned out. The solution was pretty simple, just get another power supply that puts out the same voltage but more amps(I think I went with twice the output that was marked on the wall wart). I picked up a lot of these units at the used computer store for a couple of bucks, spent $10 for a new wall wart(with the upped amp output) and they worked fine.
All systems were rebooted after the Netgear switch was inserted into the network. The systems that were capable of gigabit connections autosensed the new switch and set their port speed accordingly.
I think you should not set any port speed, just leave them in auto mode: speed auto and everything else also.
Let switch to decide if it can speed its port up to 1Gb/s. And also make sure that none of your ethernet card has only "1Gb mode". Put everything on auto.
Also you can connect your boxes one by one to the switch.
And of course, as was suggested, check power supply.
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Original Poster
Rep:
Quote:
Originally Posted by nimnull22
I think you should not set any port speed, just leave them in auto mode: speed auto and everything else also.
Not a problem. I haven't done anything to force any given speed on any of the interfaces. I was thinking of trying to force a lower speed on the Gb interfaces in an attempt to see if the switch was just conking out when it had deal with Gb connections.
Quote:
Also you can connect your boxes one by one to the switch.
And of course, as was suggested, check power supply.
That's something I was going to try out as well. I was also considering keeping the systems that currently have a maximum speed of 100Mb on the Zonet and chaining that off the Netgear unit. If that works, it'll work as a stopgap solution until I can find a beefier power wart.
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Original Poster
Rep:
Quote:
Originally Posted by lazlow
Depending on the age of the switch(not dependent on when you bought it) Netgear was shipping the same power supply on the 8 port switch as on the 5 port switch. Two or three years ago(maybe longer) this was a real and pretty common problem. Usually the excess power draw(amps) drops the voltage so far the unit just stops working, but sometimes the unit(not the wall wart) just got hot and burned out.
I actually have one of the 5-port Netgear Gb switches (obtained for my planned DMZ). I'll check the power ratings on the power warts and see if the warts are the same. If so, someone at Netgear ought to get chewed out. Shipping a power supply that doesn't let you use all the ports ought to get someone yelled at or fired.
Quote:
The solution was pretty simple, just get another power supply that puts out the same voltage but more amps(I think I went with twice the output that was marked on the wall wart). I picked up a lot of these units at the used computer store for a couple of bucks, spent $10 for a new wall wart(with the upped amp output) and they worked fine.
Where'd you get the new power warts? Would that be something my local Radio Trash might carry? I do have some old power warts from no-longer-used modems. I'll check the voltage and current ratings. Maybe I'll get lucky.
--
Rick
Updated: I just checked the power warts for the two switches. Though they are almost identical physically, electrically they were defintely different: one had 0.5A capacity and the other 1.0A. As you might have just guessed, I had the 8-port switch using the 0.5A wart. As soon as everyone gets off their computers (I do not interrupt my daughters Manga downloads. Not if I don't want to hear about it big time.) I can swap the 8-port back into the network and see what happens. And I'll be making tags for the power warts connectors to keep from using the wrong one in the future.
Last edited by rnturn; 01-01-2011 at 09:59 PM.
Reason: Updated with result of a visual inspection of the power supplies.
What usually happens is during testing they use a quality wall wart and everything works fine. Then after production starts somebody decides to save a few pennies by using a cheaper wall wart, which does not actually provide the same number of amps as the quality unit did(despite both units being rated the same). Samsung went through the same thing on capacitors a few years back. The company that made their capacitors(for years) got bought out. With the old owners they were using 12V capacitors on a 18V power supply and everything worked just fine. The new owners cut a few corners and the capacitors blew after about two years of service. Most TV repair place now have a minimum charge of $200 just to open an lcd TV. Replacing all five of the capacitors(good ones 24V) cost less than $10. It took me under an hour to replace the capacitors. You can tell the capacitors are going out when the tv starts clicking a few times before the screen comes on. Word got out that I could fix them and I wound up going through five cases of capacitors (1/2 the cost when you buy them by the case).
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Original Poster
Rep:
Quote:
Originally Posted by lazlow
I picked up a lot of these units at the used computer store for a couple of bucks, spent $10 for a new wall wart(with the upped amp output) and they worked fine.
Just out of curiousity: What amperage power warts did you wind up using? I found an old one (12V/2A) from a dead external USB drive and that didn't solve the dropped packet problem.
If the 2A is not solving the problem then a shortage of amps is not (currently) the issue. Off the top of my head I THINK the orignals were .5 amp and I went with a 1 amp.
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Original Poster
Rep:
Quote:
Originally Posted by lazlow
If the 2A is not solving the problem then a shortage of amps is not (currently) the issue. Off the top of my head I THINK the orignals were .5 amp and I went with a 1 amp.
Unless there could be networking configuration problem on one of the connected systems, I'm rapidly coming to the conclusion that the switch is a dud. The only thing that I haven't tried is to force the systems with gigabit capability to run at 100Mb. Even if that gets the switch to work, it doesn't fit in with the long term plan for the network and there's no advantage to using the Netgear unit in the network instead of the Zonet (except it'd be a short-term solution to the Zonet's sensitivity to the power glitches).
Is your Netgear switch really switch?
Does it have additional configuration through web interface or telnet?
Have you tried to connect computers with ONLY 100Mb ethernet cards?
What kind cables do you use? Old ones, straight or crossover?
If you have new short cables you can put your switch near with two boxes and connect them to switch with new short straight cables. You need to test your switch or just return it back and buy more expensive one.
Running the switch with the .5 amp wart MAY have damaged it permamently or it may have been bad out of the box. I would try forcing the machines to use GigE (sometimes switching back and forth can cause issues) and check the cables. Unless you have long runs(30m+) they usually either run GigE or not(not intermitant). If those two things do not show anything, I would return it. On GigE wether they are straight or crossover is irrelevent (the negotiation is part of the GigE standard).
Distribution: openSUSE, Raspbian, Slackware. Previous: MacOS, Red Hat, Coherent, Consensys SVR4.2, Tru64, Solaris
Posts: 2,800
Original Poster
Rep:
Quote:
Is your Netgear switch really switch?
Well, that's what the box it came in said it was. It has 9 connections: one for power and 8 RJ45 ports. No ports available for any type of "management" or configuration. I think I'd know if it was a managed switch; I'd know because of the increased cost over what I paid for this unit.
Quote:
Does it have additional configuration through web interface or telnet?
Not according to the manual (which tells you, basically, to power it up and plug in your systems. Not much else, sadly.
Quote:
Have you tried to connect computers with ONLY 100Mb ethernet cards?
At one point, yes. Things head South very quickly shortly after the first Gb-capable system is powered up and establishes a connection. I haven't figured out how to force the Gb interfaces to limit themselves to 100Mb. I tried that a while back and it didn't seem to change the interface speed.
Quote:
What kind cables do you use? Old ones, straight or crossover?
Some cables are older (not ancient) and some new. None are crossover cables. The Gb-capable systems have had their cables replaced with a variety of other cables (mixture of old and new but mostly new). Changing cables made no noticable difference in the way the switch worked. (I.e., packet loss still hit 100% in short order.)
Quote:
You need to test your switch or just return it back and buy more expensive one.
I've been testing it from time to time -- when nobody needs to use the network (which is not very often) -- and it's looking like a replacement unit is, indeed, the only thing left to try for now.
I'll have another chance to do more testing this weekend but I suspect that replacement -- with a different brand -- is the only thing that's going to help.
LinuxQuestions.org is looking for people interested in writing
Editorials, Articles, Reviews, and more. If you'd like to contribute
content, let us know.