SEO

Hi,

Our system administrator doesn't have time to look at this so I thought
I'd have a look. I'm a developer and don't know much about networks
etc. so please bear with me.

We have a 64 bit AIX 5.2 box which can't use NFS due to time out errors.
I've found that using ftp or rcp is very very slow - ftp reports about
20kb/s. This is between the AIX machine and another UNIX machine on a
100Mb/s ethernet hub.

We've got about 10 or so UNIX machines on this hub running various
versions of UNIX. We have a 32 bit AIX 4.3 machine which works fine.

I've used SMIT to look at the ethernet card. It's set to auto
negotiation which we've had problems with on other systems. Setting it
to 100Mbs doesn't help.

I tried doing an error log but nothing appeared in it other than the
start/stop messages. While the error log was running I tried to use an
NFS drive and got the usual error -

A remote host did not respond within the timeout period.
NFS write error on host sgnfs1: 78.


Sorry if this is all a bit vague - can anyone talk me through how to
investigate this?

Thanks.

Peter
--
http://www.beluga.freeserve.co.uk

Peter Smithson wrote:
> Hi,
>
> Our system administrator doesn't have time to look at this so I thought
> I'd have a look. I'm a developer and don't know much about networks
> etc. so please bear with me.
>
> We have a 64 bit AIX 5.2 box which can't use NFS due to time out errors.
> I've found that using ftp or rcp is very very slow - ftp reports about
> 20kb/s. This is between the AIX machine and another UNIX machine on a
> 100Mb/s ethernet hub.
>
> We've got about 10 or so UNIX machines on this hub running various
> versions of UNIX. We have a 32 bit AIX 4.3 machine which works fine.
>
> I've used SMIT to look at the ethernet card. It's set to auto
> negotiation which we've had problems with on other systems. Setting it
> to 100Mbs doesn't help.
>
> I tried doing an error log but nothing appeared in it other than the
> start/stop messages. While the error log was running I tried to use an
> NFS drive and got the usual error -
>
> A remote host did not respond within the timeout period.
> NFS write error on host sgnfs1: 78.

you way check in /etc/filesystems , if (on of) the mount options of this
mount point in questions is set to `soft'

if so could try to set it to `hard'

Once it helped me.
See `man mount' for details.

--
reply to usenet only - email account disabled

Peter Smithson wrote:
> Hi,
>
> Our system administrator doesn't have time to look at this so I thought
> I'd have a look. I'm a developer and don't know much about networks
> etc. so please bear with me.
>
> We have a 64 bit AIX 5.2 box which can't use NFS due to time out errors.
> I've found that using ftp or rcp is very very slow - ftp reports about
> 20kb/s. This is between the AIX machine and another UNIX machine on a
> 100Mb/s ethernet hub.
>
> We've got about 10 or so UNIX machines on this hub running various
> versions of UNIX. We have a 32 bit AIX 4.3 machine which works fine.
>
> I've used SMIT to look at the ethernet card. It's set to auto
> negotiation which we've had problems with on other systems. Setting it
> to 100Mbs doesn't help.
>
> I tried doing an error log but nothing appeared in it other than the
> start/stop messages. While the error log was running I tried to use an
> NFS drive and got the usual error -
>
> A remote host did not respond within the timeout period.
> NFS write error on host sgnfs1: 78.

you may check in /etc/filesystems , if (one of) the mount options of the
mount point in questions is set to `soft'

if so could try to set it to `hard'

It helped me once.

But see `man mount' for details before.

--
reply to usenet only - email account disabled

I'm not a networking wiz, but in our office the usual cause for slow
network response is exactly what you hit on in your message. When we
have had our Ethernet cards set to auto-negotiation, the card looks at
the network and seems to take the speed of the slowest network card,
usually an old router or hub somewhere. It often ends up being set to
10mbs Half-Duplex.

Try changing the card to 10_full_duplex or 100_full duplex.

smit devices
- Communications
- Ethernet
- Adapter
- Change/Show
Choose your adapter (ent0?)
Verify field called "Media Speed"

You may need to restart the network daemons (inetd, portmap, etc) to
get it to start using the new settings. Keep in mind that sometimes
changing these settings will stop your network connections from working.

"Peter Smithson" <pgsmithson_deletethisbit@yahoo.co.uk> wrote in message
news:MPG.1c45e5f94c90da67989820@News.Individual.NE T...
> Hi,
>
> Our system administrator doesn't have time to look at this so I thought
> I'd have a look. I'm a developer and don't know much about networks
> etc. so please bear with me.
>
> We have a 64 bit AIX 5.2 box which can't use NFS due to time out errors.
> I've found that using ftp or rcp is very very slow - ftp reports about
> 20kb/s. This is between the AIX machine and another UNIX machine on a
> 100Mb/s ethernet hub.
>
> We've got about 10 or so UNIX machines on this hub running various
> versions of UNIX. We have a 32 bit AIX 4.3 machine which works fine.
>
> I've used SMIT to look at the ethernet card. It's set to auto
> negotiation which we've had problems with on other systems. Setting it
> to 100Mbs doesn't help.
>
> I tried doing an error log but nothing appeared in it other than the
> start/stop messages. While the error log was running I tried to use an
> NFS drive and got the usual error -
>
> A remote host did not respond within the timeout period.
> NFS write error on host sgnfs1: 78.
>
>
> Sorry if this is all a bit vague - can anyone talk me through how to
> investigate this?
>
> Thanks.
>
> Peter
> --
Peter,
1. Try 'hard-setting' all your NICs instead of auto-negotiating.
2. If you're still using an ethernet hub, replace with an ethernet switch.

alvin

In article <1104929934.374004.56710@z14g2000cwz.googlegroups. com>, steven_nospam at Yahoo! Canada <steven_nospam@yahoo.ca> wrote:
> I'm not a networking wiz, but in our office the usual cause for slow
> network response is exactly what you hit on in your message. When we
> have had our Ethernet cards set to auto-negotiation, the card looks at
> the network and seems to take the speed of the slowest network card,
> usually an old router or hub somewhere. It often ends up being set to
> 10mbs Half-Duplex.
>
> Try changing the card to 10_full_duplex or 100_full duplex.

He said he was on an hub (and if it's really a hub, and not a switch),
which means 100/half duplex.

-Dan

"Peter Smithson" <pgsmithson_deletethisbit@yahoo.co.uk> schrieb im
Newsbeitrag news:MPG.1c45e5f94c90da67989820@News.Individual.NE T...
> Hi,
>
> Our system administrator doesn't have time to look at this so I thought
> I'd have a look. I'm a developer and don't know much about networks
> etc. so please bear with me.
>
> We have a 64 bit AIX 5.2 box which can't use NFS due to time out errors.
> I've found that using ftp or rcp is very very slow - ftp reports about
> 20kb/s. This is between the AIX machine and another UNIX machine on a
> 100Mb/s ethernet hub.
>
> We've got about 10 or so UNIX machines on this hub running various
> versions of UNIX. We have a 32 bit AIX 4.3 machine which works fine.
>
> I've used SMIT to look at the ethernet card. It's set to auto
> negotiation which we've had problems with on other systems. Setting it
> to 100Mbs doesn't help.
>
> I tried doing an error log but nothing appeared in it other than the
> start/stop messages. While the error log was running I tried to use an
> NFS drive and got the usual error -
>
> A remote host did not respond within the timeout period.
> NFS write error on host sgnfs1: 78.
>
>
> Sorry if this is all a bit vague - can anyone talk me through how to
> investigate this?
>
> Thanks.
>
> Peter
> --

Hallo Peter,

You could use
# entstat -d ent0
to see (among other usefull stuff) collision statistics and media speed
running. Make sure that there are enough nfs daemons running on the nfs
server. And (as the other posters suggested) if you can avoid the auto
negotiation as AIX was never good with that (and still is not).

HTH,
Andreas

In article <crgrno$k2h$1@tgx093.str.allianz.de>, b79xan@gmx.de says...

>
> You could use
> # entstat -d ent0
> to see (among other usefull stuff) collision statistics and media speed
> running. Make sure that there are enough nfs daemons running on the nfs
> server. And (as the other posters suggested) if you can avoid the auto
> negotiation as AIX was never good with that (and still is not).

I'm not too worried about NFS specific problems as I think NFS isn't
working due to a general network problem - otherwise ftp & rcp would be
OK.

I've used that command you gave and I think it's telling me something -

------------------------------------------------------------------
Elapsed Time: 1 days 1 hours 13 minutes 57 seconds

Transmit Statistics: Receive Statistics:
-------------------- -------------------
Packets: 152693 Packets: 327703
Bytes: 55840399 Bytes: 115845744
Interrupts: 488 Interrupts: 324792
Transmit Errors: 1856 Receive Errors: 0
Packets Dropped: 1 Packets Dropped: 0
Bad Packets: 0
Max Packets on S/W Transmit Queue: 86
------------------------------------------------------------------

There are quite a lot of Transmit errors for 1 day's usage (no recieve
errors). I've noticed that NFS works fine as long as I'm only reading
from the NFS drive.

The same command on our AIX 4 box shows -

------------------------------------------------------------------
Elapsed Time: 0 days 7 hours 34 minutes 5 seconds

Transmit Statistics: Receive Statistics:
-------------------- -------------------
Packets: 83505 Packets: 125555
Bytes: 40863328 Bytes: 43194960
Interrupts: 83501 Interrupts: 125266
Transmit Errors: 0 Receive Errors: 160
Packets Dropped: 0 Packets Dropped: 0
Bad Packets: 0
Max Packets on S/W Transmit Queue: 9
------------------------------------------------------------------

No trasmit errors there at all.

I think the auto negotiate is OK -

Media Speed Running: 100 Mbps Half Duplex

Someone asked if it was really a "hub". I've spoken to our sysadmin and
he says our UNIX network is on it's own "segment" - there is no hub or
switch. But then he said that you could say they all use a 10/100
switch - not sure what all that means. I guess it means they're on a
switch. There is a bridge to our Windows network.

Another difference between this and the AIX 4 is that in the "Driver
flags" section the AIX 5 box (the one with the problem) has
"PrivateSegment DataRateSet". This could just be a new feature of the
OS - there was quite a lot more output on the AIX 5 machine.

Thanks for all the replies.

Peter

--
http://www.beluga.freeserve.co.uk

Peter Smithson wrote:
> In article <crgrno$k2h$1@tgx093.str.allianz.de>, b79xan@gmx.de says...
>
>
>>You could use
>># entstat -d ent0
>>to see (among other usefull stuff) collision statistics and media speed
>>running. Make sure that there are enough nfs daemons running on the nfs
>>server. And (as the other posters suggested) if you can avoid the auto
>>negotiation as AIX was never good with that (and still is not).
>
>
> I'm not too worried about NFS specific problems as I think NFS isn't
> working due to a general network problem - otherwise ftp & rcp would be
> OK.
>
> I've used that command you gave and I think it's telling me something -
>
> ------------------------------------------------------------------
> Elapsed Time: 1 days 1 hours 13 minutes 57 seconds
>
> Transmit Statistics: Receive Statistics:
> -------------------- -------------------
> Packets: 152693 Packets: 327703
> Bytes: 55840399 Bytes: 115845744
> Interrupts: 488 Interrupts: 324792
> Transmit Errors: 1856 Receive Errors: 0
> Packets Dropped: 1 Packets Dropped: 0
> Bad Packets: 0
> Max Packets on S/W Transmit Queue: 86
> ------------------------------------------------------------------
>
> There are quite a lot of Transmit errors for 1 day's usage (no recieve
> errors). I've noticed that NFS works fine as long as I'm only reading
> from the NFS drive.
>
> The same command on our AIX 4 box shows -
>
> ------------------------------------------------------------------
> Elapsed Time: 0 days 7 hours 34 minutes 5 seconds
>
> Transmit Statistics: Receive Statistics:
> -------------------- -------------------
> Packets: 83505 Packets: 125555
> Bytes: 40863328 Bytes: 43194960
> Interrupts: 83501 Interrupts: 125266
> Transmit Errors: 0 Receive Errors: 160
> Packets Dropped: 0 Packets Dropped: 0
> Bad Packets: 0
> Max Packets on S/W Transmit Queue: 9
> ------------------------------------------------------------------
>
> No trasmit errors there at all.
>
> I think the auto negotiate is OK -
>
> Media Speed Running: 100 Mbps Half Duplex


IF you have the switch in either auto or full duplex normally the
adapter will negociate to half. So, if you want to avoid problems, set
BOTH the adapter and the switch to 100/Full Duplex. Also, check the CAT
of the cable. It must be at least 5.

--
BC

Peter Smithson <pgsmithson_deletethisbit@yahoo.co.uk> wrote in
news:MPG.1c463d4138425ee4989821@News.Individual.NE T:
> [...]
> Transmit Statistics: Receive Statistics:
> -------------------- -------------------
> Packets: 83505 Packets: 125555
> Bytes: 40863328 Bytes: 43194960
> Interrupts: 83501 Interrupts: 125266
> Transmit Errors: 0 Receive Errors: 160
160/125555 is less than 0.1%, not really serious, but still too high. If
the cabling is correct and the network equipment wasn't bough on Toys'R
US, then you should see zero to two receive errors for about a billion
(american) received packets.


> Packets Dropped: 0 Packets Dropped: 0
> Bad Packets: 0
> Max Packets on S/W Transmit Queue: 9
> ------------------------------------------------------------------
>
> No trasmit errors there at all.
>
> I think the auto negotiate is OK -
>
> Media Speed Running: 100 Mbps Half Duplex
>
> Someone asked if it was really a "hub". I've spoken to our sysadmin and
> he says our UNIX network is on it's own "segment" - there is no hub or
> switch. But then he said that you could say they all use a 10/100
> switch - not sure what all that means.
If he said *that*, it means he's a clueless fuck. If it's ethernet it's
either: a hub, a switch, or a direct cable between two machines. Are you
*sure* he said *that* ?

A segment doesn't exist out of thin air. It has a physical reality that
is one of the three previous things.

> I guess it means they're on a switch.
Just ask him again.

> There is a bridge to our Windows network.
That's cruel. Windows should be segregated behind a router.

> Another difference between this and the AIX 4 is that in the "Driver
> flags" section the AIX 5 box (the one with the problem) has
> "PrivateSegment DataRateSet". This could just be a new feature of the
> OS - there was quite a lot more output on the AIX 5 machine.
Just ignore it. Check to see if you have CRC errors or large numbers of
collisions.

CRC errors and incomplete packets are quite common when you force the
card to 100 Full and the other end is a hub (or a switch forced to 100
Half).

Large numbers of collisions are quite common when you force the card to
100 Half and the other end is a switch forced to 100 Full.