Mailing List Archive

Simon,
Did you change anything on your network recently? Are you doing auto
negotiation on the Ethernet ports on the Netapp? If so, force them and the
switch port to 100Mbit full duplex (unless you are using Gigabit Ethernet
which you will leave as Auto negotiate). Run "ifstat -a" on the Netapp and
see if you have excessive errors/collisions on a interface. Run tracert from
the machines having slow access to the Netapp and see if they are going
through any unusual hops.
Mike


-----Original Message-----
From: Clawson, Simon [mailto:simon_clawson@mentorg.com]
Sent: Monday, October 28, 2002 9:43 AM
To: 'Toasters@mathworks.com'
Subject: CIFS access slow


Hi,

I have sen over the last few days a marked decrease in CIFS performance from
our 760. We have a shell for the PC (TCSH) and if I perform an ls -l on a
top level share mapped to a network drive on the PC I get a 30-40 second
delay before the list is returned.

I saw several lsa lookup errors last week, but at the moment these are all
clear - but still the performance is poor. We use UNIX style qtrees, and
have both WINS and NIS services.

Any ideas?

Simon

Simon Clawson TEL:- +44
(0)1635 811409
HDL Designer Series Team Systems Administrator FAX:- +44 (0)1635
810108
Mentor Graphics Ireland Ltd (UK Branch) MOB:- +44
(0)7788 716071
Rivergate
London Road
Newbury
Berkshire
RG14 2QB

In addition to Mike's suggestion, you can give the sysstat command a go.

sysstat -s -x 2

This will summarise the output of sysstat and print out its output every 2
seconds. I guess the fields that would be most relevant in this case are

1. CPU
- heavy NFS, CIFS, restores and backups will see your CPU utilisation rise.

2. Disk KB/s
read and writes - take note of how much writes are taking place. Continous
writing would suggest a bottleneck. NVRAM should catch most of the writes
and flush it out every 10 seconds or when the the bank is coming close to
full. If continous writing (or flushing) of NVRAM is constant, you ~may~
need to add more NVRAM. Disclaimer : This is not a sales pitch -- I am not
an employee of Network Appliance :)

The Net KB/s and Cache Age field might give you a hint as well ... Is it
possible to get the output of sysstat on your filer?


-- Clarence.


On Tue, 29 Oct 2002 07:07 am, in a distant galaxy, Mike Ball, wrote:
> Simon,
> Did you change anything on your network recently? Are you doing auto
> negotiation on the Ethernet ports on the Netapp? If so, force them and the
> switch port to 100Mbit full duplex (unless you are using Gigabit Ethernet
> which you will leave as Auto negotiate). Run "ifstat -a" on the Netapp and
> see if you have excessive errors/collisions on a interface. Run tracert
> from the machines having slow access to the Netapp and see if they are
> going through any unusual hops.
> Mike
>
>
> -----Original Message-----
> From: Clawson, Simon [mailto:simon_clawson@mentorg.com]
> Sent: Monday, October 28, 2002 9:43 AM
> To: 'Toasters@mathworks.com'
> Subject: CIFS access slow
>
>
> Hi,
>
> I have sen over the last few days a marked decrease in CIFS performance
> from our 760. We have a shell for the PC (TCSH) and if I perform an ls -l
> on a top level share mapped to a network drive on the PC I get a 30-40
> second delay before the list is returned.
>
> I saw several lsa lookup errors last week, but at the moment these are all
> clear - but still the performance is poor. We use UNIX style qtrees, and
> have both WINS and NIS services.
>
> Any ideas?
>
> Simon
>
> Simon Clawson TEL:- +44
> (0)1635 811409
> HDL Designer Series Team Systems Administrator FAX:- +44 (0)1635
> 810108
> Mentor Graphics Ireland Ltd (UK Branch) MOB:- +44
> (0)7788 716071
> Rivergate
> London Road
> Newbury
> Berkshire
> RG14 2QB