Prev: showrev -p broke after Solaris 10 upgrade
Next: Solaris 10 specifics
From: Tom on 14 Mar 2007 09:17
Hi,

I'm running Solaris 10 and have a question about the
ip_multidata_outbound tunable parameter.

All the information I can find says that you turn this off with the
following command:
# ndd -set /dev/ip ip_multidata_outbound 0

But nowhere do I read how to make this change permanent. I thought I
could add the following to /etc/system so that the change would
survive reboots:
set ip:ip_multidata_outbound = 0

However, after rebooting the node I checked (with -get) and the value
had not changed.

Does anyone know how to set this in /etc/system? Is my problem syntax,
or for some reason is this parameter not setable using this mechanism?

Thank you.

From: James Carlson on 14 Mar 2007 10:04
"Tom" <cube_384(a)yahoo.com> writes:
> All the information I can find says that you turn this off with the
> following command:
> # ndd -set /dev/ip ip_multidata_outbound 0

Yes.

> But nowhere do I read how to make this change permanent. I thought I
> could add the following to /etc/system so that the change would
> survive reboots:
> set ip:ip_multidata_outbound = 0

Correct; you generally cannot set ndd parameters via /etc/system.

ndd parameters are normally held inside a data structure, and the
names are strings (rather than symbols). /etc/system knows only about
simple variables.

> However, after rebooting the node I checked (with -get) and the value
> had not changed.
>
> Does anyone know how to set this in /etc/system? Is my problem syntax,
> or for some reason is this parameter not setable using this mechanism?

Use a start-up (rc*.d) script or SMF service to run ndd.

Why do you want to do this?

--
James Carlson, Solaris Networking <james.d.carlson(a)sun.com>
Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677
From: Tom on 14 Mar 2007 11:11
On Mar 14, 8:04 am, James Carlson <james.d.carl...(a)sun.com> wrote:
> Use a start-up (rc*.d) script or SMF service to run ndd.
>
> Why do you want to do this?

Thanks for the information. We have one filer that seems to give our
S10 units fits; we get a lot of fragmentations and retransmits when
doing, for example, an NFS copy from it to the S10 unit. We want to
see if this removes that problem.

I think its a problem with the filer but I'm not able to get access to
it so I'm trying to work it from this other end.

Tom

From: James Carlson on 15 Mar 2007 10:15
"Tom" <cube_384(a)yahoo.com> writes:
> On Mar 14, 8:04 am, James Carlson <james.d.carl...(a)sun.com> wrote:
> > Use a start-up (rc*.d) script or SMF service to run ndd.
> >
> > Why do you want to do this?
>
> Thanks for the information. We have one filer that seems to give our
> S10 units fits; we get a lot of fragmentations and retransmits when
> doing, for example, an NFS copy from it to the S10 unit. We want to
> see if this removes that problem.

Ah, ok. Any time there's rotten network performance, I'd suggest
looking at the Ethernet itself. Make sure you're not 'forcing' the
duplex mode -- doing so typically causes the peer to fall into
half-duplex, because of the way the 802 standards work.

> I think its a problem with the filer but I'm not able to get access to
> it so I'm trying to work it from this other end.

That makes some sense. As long as you're doing this, another thing
I'd poke around with here would be checksum offload; 'set ip:dohwcksum
= 0' in /etc/system.

However, rather than hacking around at the system, I'd suggest that a
better course would be to gather snoop traces and statistics (such as
'netstat -s' and kstats on the network device), and calling Sun's
support for help.

--
James Carlson, Solaris Networking <james.d.carlson(a)sun.com>
Sun Microsystems / 1 Network Drive 71.232W Vox +1 781 442 2084
MS UBUR02-212 / Burlington MA 01803-2757 42.496N Fax +1 781 442 1677
 | 
Pages: 1
Prev: showrev -p broke after Solaris 10 upgrade
Next: Solaris 10 specifics