Re: network failures

From: Bill Vermillion (bv_at_wjv.com)
Date: 02/20/05 

Date: Sun, 20 Feb 2005 04:35:12 GMT

In article <111eu9pat361993@corp.supernews.com>,
news.nuvox.net <fabiog@venmar.com> wrote:
>
>"J.Salfer" <JohnS@VoyageurSoftware.com> wrote in message
>news:421651af$1_1@newspeer2.tds.net...
>> "Ronald J Marchand" <ron@rojomar.com> wrote in message
>> news:f2e1c$42164df8$42a6716f$28713@msgid.meganewsservers.com...
>>> "J.Salfer" <JohnS@VoyageurSoftware.com> wrote in message
>>> news:42164622$1_1@newspeer2.tds.net...
>>>> Hello,
>>>>
>>>> Recently our SCO OSR505 box has started having network problems where we
>>>> loose all connections and cannot reconnect. The system is rebooted to
>>>> get
>>>> things working again.
>>>>
>>>> When netstat -m is viewed we have huge failures in class 6 and in this
>>> case
>>>> class 7 lines. We've increased the NSTRPAGES from 1400 to 3500. That
>>> seemed
>>>> to reduce the frequency, but the problem still comes up.
>>>>
>>>> We could keep increasing the value, but I'm wondering what all the
>>>> sudden
>>>> started causing the problem as the system has been stable for 3 years
>>> prior
>>>> to this coming up.
>>>>
>>>> Comments would be greatly appreciated.
>>>>
>>>> Thanks,
>>>> John.
>>>
>>> Had a similar situation with 5.0.6 and a 3Com card. Worked well when
>>> installed. Over time the number of connections and the traffic on the
>>> system increased. Replaced the card with an Intel Pro 1000 and all has
>>> been
>>> well since.
>>>
>>> Ron
>>>
>>>
>>
>> Interesting. The machine has an Intel Pro100 on-board NIC that we disabled
>> to installed a 3Com early in our troubleshooting. Our switch was stating
>> we were having a low number of bad packets being sent out and we thought
>> maybe the problems were related. The hang problem continued though and it
>> brought us to find the failures and that increasing NSTRPAGES helped some.
>> Makes me wonder if I should switch back to the on-board card.
>>
>> We increased NSTRPAGES to 8000 for now to see if it'll resolve the issue.

>> I think I should still find something to monitor incoming
>> packets. Thinking that maybe something on the network has
>> starting hitting the system being the problem came up out of
>> the blue.

>> Anyone know of a good program? So far I've come across nmap,
>> tcpdump, snoop and etherfind, but I'm not familiar with any of
>> them so I don't know if one is better than the other.

>> John.

>May or may not be an issue, but in the past, letting both the
>card and the switch auto-negotiate speed and duplex has caused
>problems. Hard coding both (or just one if you can't do both) to
>the appropriate settings (i.e 100Mbs/full-duplex) might help.

Actually you have to let both auto-negotiate. If auto-negotiation
fails you will find it defaulting to a mode which is usually not
good.

For a very good discussion on how things can go wrong and examples
of different settings, see this link.

http://www.cisco.com/warp/public/473/46.html

It also points out that major problems occur when some vendors
decided to add enhancements that aren't correctly specified.

It's a longish document - but it is very good.

Bill 

-- 
Bill Vermillion - bv @ wjv . com

Relevant Pages

  • Re: custom install cd
    ... locks up that switch in seconds and it must be power cycled. ... Next time do not turn the DL380 off but disconnect the wallpower to ... If so you have a grounding problem. ... Bill Vermillion - bv @ wjv. ...
    (comp.unix.sco.misc)
  • Re: DE500 - DE602 and auto-negociate mode
    ... >> with new network equipement. ... > I have found that it is always best to NOT rely on auto-negotiate. ... If the auto-negotiate fails and you get a mismatch between your switch ...
    (comp.os.vms)
  • Re: DE500, Decnet, VMS 6.2 on Alpha 800
    ... >indicating that the Tulip based cards had issues with auto-negotiate ... but the switch LEDs say half-duplex and performance is ...
    (comp.os.vms)
  • Re: Slow boot up
    ... Macintos workstations and some Windows platform workstations are in a same ... correct that Macs work better when both sides are left to auto-negotiate. ... allows the switch and computer to negotiate speed as well as duplex. ...
    (microsoft.public.macintosh.general)
  • Re: Odd network issue ... *very* slow scp between two servers
    ... When I did a traceroute to the adjacent machine which was on a ... separate /24 network, the packets went to the switch, to the ... intervening bridge - post switch / pre router that added delay. ... Bill Vermillion - bv @ wjv. ...
    (freebsd-stable)