f



networking problem (very low networking bandwith after power failure of switch)

Dear list members,

I have a strange networking problem with a solaris 8 mashine :

This mashine has a very low network bandwith ( needs several hours
to copy 1 Gig of data through a 100 Mbit network, over a switch) using
scp.

A possible cause of that problem could be, that the networking switch
to which the mashine is connected powered down due to a power supply 
failure.
After powering up that switch again I had no problems accessing the 
solaris box again. But the connection is very slow.

First I thought that I could have a problem with the nfs-client on that 
mashine causing to much traffic since it has to import several dozens of 
directories. But the problem wasn't fixed by shutting down the nfs-client.

I also checked the ndd-switches but they were ok.

My question :
How could I check what the cause of my problem is ?

Did you have similar problems and how could you fix them ?

Best regards,

Daniel
0
Daniel
8/31/2005 11:32:03 AM
comp.sys.sun.admin 3740 articles. 0 followers. bozothedeathmachine16 (49) is leader. Post Follow

4 Replies
949 Views

Similar Articles

[PageSpeed] 35

In article <df44g0$gn5$1@newsreader3.netcologne.de>,
Daniel Wetzler  <danielwetzler@appleinfo.de> wrote:

> I have a strange networking problem with a solaris 8 mashine :

First thing that I would check for would be a duplex miss-match
between the switch and the Solaris box.

Regards,
-- 
Neil Hoggarth                                Departmental Computing Manager
<neil.hoggarth@physiol.ox.ac.uk>                   Laboratory of Physiology
http://www.physiol.ox.ac.uk/~njh/                  University of Oxford, UK
0
njh
8/31/2005 12:39:50 PM
Daniel Wetzler wrote:
> Dear list members,
> 
> I have a strange networking problem with a solaris 8 mashine :
> 
> This mashine has a very low network bandwith ( needs several hours
> to copy 1 Gig of data through a 100 Mbit network, over a switch) using
> scp.
> 
> A possible cause of that problem could be, that the networking switch
> to which the mashine is connected powered down due to a power supply 
> failure.
> After powering up that switch again I had no problems accessing the 
> solaris box again. But the connection is very slow.
> 
> First I thought that I could have a problem with the nfs-client on that 
> mashine causing to much traffic since it has to import several dozens of 
> directories. But the problem wasn't fixed by shutting down the nfs-client.
> 
> I also checked the ndd-switches but they were ok.
> 
> My question :
> How could I check what the cause of my problem is ?
> 
> Did you have similar problems and how could you fix them ?
> 
> Best regards,
> 
> Daniel

- check with "iostat -e" and "nfsstat" if you are getting errors on your nfs.
- check with "netstat -I <interface>" if you are getting errors on your network interface
if you are getting errors, check the interface settings, cables, and try different ports on your switch and possible 
another nic in your computer

HTH,
Tom
0
Thomas
8/31/2005 2:44:10 PM
Daniel Wetzler wrote:

> First I thought that I could have a problem with the nfs-client on that 
> mashine causing to much traffic since it has to import several dozens of 
> directories. But the problem wasn't fixed by shutting down the nfs-client.
> 
> I also checked the ndd-switches but they were ok.
> 
> My question :
> How could I check what the cause of my problem is ?
> 
> Did you have similar problems and how could you fix them ?

   My first guest would be the duplex problems as Neil already 
suggested.  This would be especially true if:
   - you were seeing about 10% of the bandwidth that you should have
   - everything else was dreadfully slow as well (try ftp)

   We've run into the slow-nfs problem once with Solaris8 and some old 
version of Linux though, where the problem was some kind of Linux 
nfs-server implementation.  There was nothing really to do about it 
other than upgrade.

   Are you using UDP or TCP for NFS?  If using UDP, perhaps you could 
try to use TCP, and see if it helps anything?

   /Marcin
0
Marcin
9/5/2005 11:31:52 AM
On Mon, 05 Sep 2005 11:31:52 GMT, Marcin Dobrucki
<Marcin.Dobrucki@TAKETHISAWAY.nokia.com> wrote:

>Daniel Wetzler wrote:
>
>> First I thought that I could have a problem with the nfs-client on that 
>> mashine causing to much traffic since it has to import several dozens of 
>> directories. But the problem wasn't fixed by shutting down the nfs-client.
>> 
>> I also checked the ndd-switches but they were ok.
>> 
>> My question :
>> How could I check what the cause of my problem is ?
>> 
>> Did you have similar problems and how could you fix them ?
>
>   My first guest would be the duplex problems as Neil already 
>suggested.  This would be especially true if:
>   - you were seeing about 10% of the bandwidth that you should have
>   - everything else was dreadfully slow as well (try ftp)
>
>   We've run into the slow-nfs problem once with Solaris8 and some old 
>version of Linux though, where the problem was some kind of Linux 
>nfs-server implementation.  There was nothing really to do about it 
>other than upgrade.
>
>   Are you using UDP or TCP for NFS?  If using UDP, perhaps you could 
>try to use TCP, and see if it helps anything?
>
>   /Marcin

... you've checked everything but the hardware itself!
If you havent' changed anything, and you've gone through cold boots
before without any problem, and your hardware doesn't work anymore,
then consider that maybe the hardware is broken, and replace the NIC.

(The same thing happened to me, but my NIC was on the motherboard, I
bought a replacement add-on that I could plug into a free PCI slot and
things worked fine.  I picked up a NIC for very little money from a
broker selling on ebay).

Remember, hardware does break ;)

0
David
9/5/2005 2:55:48 PM
Reply: