half or full duplex?
Legend
Is it true that all the Ultra machine's NIC cards are capble of running full
duplex? If not, how can I find out whether a NIC card is able to handle full
duplex mode?
To enable full duplex, is it just add these line to /etc/system and restart?
set hme:hme_adv_autoneg_cap=0
set hme:hme_adv_100hdx_cap=0
set hme:hme_adv_100fdx_cap=1
Please correct me if I am wrong. The concept of full duplex is allow in and
out communication through the NIC card simultaneously?
Regards,
Benny
SR
The /etc/system entries force full duplex. Unless you have a very stable
network (i.e. switches that never go down, I'd leave it alone).
Stew the SailBum.com
"Legend" <leg...@spacelab.net> wrote in message
news:T8qg4.569$Ve7....@iad-read.news.verio.net...
Robert Yoder
>
> Hello all,
>
> Is it true that all the Ultra machine's NIC cards are capble of running full
> duplex? If not, how can I find out whether a NIC card is able to handle full
> duplex mode?
If your network device is named "hme*" or "qfe*"
it is a 10/100 device.
> To enable full duplex, is it just add these line to /etc/system and restart?
> set hme:hme_adv_autoneg_cap=0
> set hme:hme_adv_100hdx_cap=0
> set hme:hme_adv_100fdx_cap=1
_IF_ the hub/switch you are plugging into supports auto-negotiation,
you don't need to do anything.
otherwise:
_IF_ the hub/switch does NOT do auto-negotiation
AND is capable of full duplex
_THEN_ Solaris will fall back to half duplex
_UNLESS_ you force it full duplex
You can force it to full duplex by 3 different methods:
a) Just what you described.
b) Modifying /kernel/drv/hme.conf.
c) Putting "ndd" commands into a boot time startup script.
ry
--
ryo...@dimensional.com
"Unix: The Solution to the W2K Problem."
Alberto da Silva
recommended.
The hme interface was developed before that full duplex / half duplex
standard was defined.
If you force your switch to full duplex, hme won't detect this and operates
in half duplex.
We use a number of Sun (new and old) connected to Cisco CAT 5xxx, and even
with the latest flash, etc, these Suns (and compaqs, dells, etc) don't
detect the forced full
duplex on the switch side.
When you get this misconfiguration, performance drops off dramatically (lost
packets, etc).
I recommend forcing both sides to full duplex.
As a rule we force all out Cat5000 100Mb/s switched ports to full duplex.
Alberto
mailto:Alberto...@rmb.co.za.nospam
"SR" <St...@nospam.sailbum.com> wrote in message
news:s84u1h...@corp.supernews.com...
> Most Ultra-class built-in NICs autosense half/full duplex mode.
> The /etc/system entries force full duplex. Unless you have a very stable
> network (i.e. switches that never go down, I'd leave it alone).
>
> Stew the SailBum.com
> "Legend" <leg...@spacelab.net> wrote in message
> news:T8qg4.569$Ve7....@iad-read.news.verio.net...
> > Hello all,
> >
> > Is it true that all the Ultra machine's NIC cards are capble of running
> full
> > duplex? If not, how can I find out whether a NIC card is able to handle
> full
> > duplex mode?
> >
> > To enable full duplex, is it just add these line to /etc/system and
> restart?
> > set hme:hme_adv_autoneg_cap=0
> > set hme:hme_adv_100hdx_cap=0
> > set hme:hme_adv_100fdx_cap=1
> >
Casper H.S. Dik - Network Security Engineer
"Alberto da Silva" <Alberto...@rmb.co.za> writes:
>Unfortunately leaving Solaris to autodetect the full / half duplex is NOT
>recommended.
That's not true; it works fine module poatches for 2.5.1/2.5. Forcing it
will stop it negotiating altogether which may gives trouble at the switch
end. If the switch is fixed, it too won't negotiate (and Solaris will
fall back to half duplex).
With current switch firmware and Solaris software, auto negotiation is
the best you can do.
>The hme interface was developed before that full duplex / half duplex
>standard was defined.
>If you force your switch to full duplex, hme won't detect this and operates
>in half duplex.
Yes, that's because the switch fails to negotiate in such cases; it's not
a failure of the Sun interface.
>We use a number of Sun (new and old) connected to Cisco CAT 5xxx, and even
>with the latest flash, etc, these Suns (and compaqs, dells, etc) don't
>detect the forced full duplex on the switch side.
That's because you shouldn't force the switch.
When you force the switch the switch will not negotiate.
(Or do you blame compaq/dell and Sun for the same problem).
No, you need to do yourself a favour and switch the switches to
auto detect. (Although you will need to switch all yoru systems
back to auto-detect by hand too)
Casper
--
Expressed in this posting are my opinions. They are in no way related
to opinions held by my employer, Sun Microsystems.
Statements on Sun products included here are not gospel and may
be fiction rather than truth.
Lon Stowell
Casper H.S. Dik - Network Security Engineer <Caspe...@Holland.Sun.Com> wrote:
>(Or do you blame compaq/dell and Sun for the same problem).
Who to blame? Who cares? The resulting rotten thruput
when there is a duplex mismatch is a very significant
problem.
>
>No, you need to do yourself a favour and switch the switches to
>auto detect. (Although you will need to switch all yoru systems
>back to auto-detect by hand too)
If you DO follow this advice, test the thruput. Any sun
can saturate 100 mbit easily to another sun, particularly
a U-10 or faster. The NFS timeouts, poor file transfer
times, just aren't worth the risk of trusting possibly
buggy switches and/or NIC's....no matter who is at fault.
Lon Stowell
>
>_IF_ the hub/switch you are plugging into supports auto-negotiation,
>you don't need to do anything.
Unduly optimistic. IF the switch supports auto-negotiation
AND can perform this successfully with the QFE/HME, you
don't need to do anything. Until the phase of the moon
changes. Some can autonegotiate, disturbingly many can't.
Number one cause of poor thruput, don't risk it.
Casper H.S. Dik - Network Security Engineer
lsto...@triton.dnai.com (Lon Stowell) writes:
> If you DO follow this advice, test the thruput. Any sun
> can saturate 100 mbit easily to another sun, particularly
> a U-10 or faster. The NFS timeouts, poor file transfer
> times, just aren't worth the risk of trusting possibly
> buggy switches and/or NIC's....no matter who is at fault.
Both the Suns and switches will tell what their ports are running at;
I've seen more problems because of fixed speed settings in switches
(which are a sure way of forcing a mismatch) than as the result of
auto negotiation failures.
Casper H.S. Dik - Network Security Engineer
lsto...@triton.dnai.com (Lon Stowell) writes:
> Unduly optimistic. IF the switch supports auto-negotiation
> AND can perform this successfully with the QFE/HME, you
> don't need to do anything. Until the phase of the moon
> changes. Some can autonegotiate, disturbingly many can't.
> Number one cause of poor thruput, don't risk it.
Buy better switches or upgrade the firmware in the switches.
Robert Yoder
>
> Robert Yoder <ryo...@dimensional.com> wrote:
> >
> >_IF_ the hub/switch you are plugging into supports auto-negotiation,
>
> Unduly optimistic. IF the switch supports auto-negotiation
> AND can perform this successfully with the QFE/HME, you
> don't need to do anything. Until the phase of the moon
> changes. Some can autonegotiate, disturbingly many can't.
> Number one cause of poor thruput, don't risk it.
I have about 30 Suns with hme/qfe interfaces, and I
have NEVER seen a problem with autonegotiation.
I have heard tales of autoneg problems,
but I have never seen it.
dwm...@my-deja.com
Robert Yoder <ryo...@dimensional.com> wrote:
> Legend wrote:
> >
> > Hello all,
> >
> > Is it true that all the Ultra machine's NIC cards are capble of
running full
> > duplex? If not, how can I find out whether a NIC card is able to
handle full
> > duplex mode?
>
> it is a 10/100 device.
>
restart?
> > set hme:hme_adv_autoneg_cap=0
> > set hme:hme_adv_100hdx_cap=0
> > set hme:hme_adv_100fdx_cap=1
>
> you don't need to do anything.
>
>
> _IF_ the hub/switch does NOT do auto-negotiation
> AND is capable of full duplex
> _THEN_ Solaris will fall back to half duplex
> _UNLESS_ you force it full duplex
>
> You can force it to full duplex by 3 different methods:
>
> a) Just what you described.
>
> b) Modifying /kernel/drv/hme.conf.
>
> c) Putting "ndd" commands into a boot time startup script.
>
I've had no luck with (a) or (b); and method (c) results in a
momentary loss of connectivity after bootup. But (c) works, at least.
My experience with autodetect is no better than Alberto Da Silva's;
i.e. it doesn't autodetect.
David.
Sent via Deja.com http://www.deja.com/
Before you buy.
dwm...@my-deja.com
Caspe...@Holland.Sun.Com (Casper H.S. Dik - Network Security
Engineer) wrote:
> [[ PLEASE DON'T SEND ME EMAIL COPIES OF POSTINGS ]]
>
> lsto...@triton.dnai.com (Lon Stowell) writes:
>
> > AND can perform this successfully with the QFE/HME, you
> > don't need to do anything. Until the phase of the moon
> > changes. Some can autonegotiate, disturbingly many can't.
> > Number one cause of poor thruput, don't risk it.
>
>
Or you could switch to HP 9000 :>
Jonathan Magid
Robert Yoder <ryo...@dimensional.com> wrote:
>
>If your network device is named "hme*" or "qfe*"
>it is a 10/100 device.
What about be* devices? I've got a SS1000 and a SS1000E, both of
which are running at half duplex and have be* ethernet cards. The
UE4000 that's on the same switch correctly detects the network's
capability for full duplex communications.
Can I force the be* devices to run full duplex? If so, how?
Also, what command can I use to find out whether an interface is running
full or half duplex under Solaris 2.7?
thanks,
jem.
--
j...@metalab.unc.edu|SEEKING: High level opening in New World Order.
| Assigned black helicopter is a must.
C. Newport
: In article <3882A45D...@dimensional.com>,
: Robert Yoder <ryo...@dimensional.com> wrote:
: >
: >If your network device is named "hme*" or "qfe*"
: >it is a 10/100 device.
: [suggestions deleted]
: What about be* devices? I've got a SS1000 and a SS1000E, both of
: which are running at half duplex and have be* ethernet cards. The
: UE4000 that's on the same switch correctly detects the network's
: capability for full duplex communications.
: Can I force the be* devices to run full duplex? If so, how?
: Also, what command can I use to find out whether an interface is running
: full or half duplex under Solaris 2.7?
Solaris 7 does not support the be device.
( according to the release notes ).
Casper H.S. Dik - Network Security Engineer
j...@triphop.metalab.unc.edu (Jonathan Magid) writes:
>What about be* devices? I've got a SS1000 and a SS1000E, both of
>which are running at half duplex and have be* ethernet cards. The
>UE4000 that's on the same switch correctly detects the network's
>capability for full duplex communications.
be is 10/100 but only runs half-duplex.
>Also, what command can I use to find out whether an interface is running
>full or half duplex under Solaris 2.7?
ndd /dev/XXX link_mode (0 for half, 1 for full).
May only work for hme; le devices cannot run full-duplex so there's no
way to ask the device how it's doing.
Ulrich Habel
(in message <3882A45D...@dimensional.com>):
>> restart?
>> set hme:hme_adv_autoneg_cap=0
>> set hme:hme_adv_100hdx_cap=0
>> set hme:hme_adv_100fdx_cap=1
>
SUN recommends putting the ndd commands in a startup script.
If you run the commands when the sun is in runlevel 2/3 will
cause an interupt in network services.
Dirk Hufnagel
>
> dwm...@my-deja.com wrote:
> >>
> >> Buy better switches or upgrade the firmware in the switches.
> >>
>
> > Or you could switch to HP 9000 :>
>
> do the same thing on my HP's as I've done to the suns --
> lock the speed and duplex. Nobody seems to play well
> with Cisco switches...
>
I connected an Ultra5 and a Sparc5 (with hme card) to
both Cisco and HP switches and never hand to do anything,
it just worked perfectly with 100Mbit/full duplex. The
switches were a Cisco Catalyst 2900 and an HP 1600M.
The SUN's were previously connected to an 10Mbit hub
and I just unplugged the cables and connected them to
the Cisco switch without even shutting down the computer.
I lost a few packets and one Xemacs crashed, but the link
came back after about 20seconds with 100Mbit/full duplex.
I did the same after byuing the HP switch (the Cisco switch
was borrowed).
BTW, Intel NIC's in PC's worked fine too, although I had
problems with some clone NIC's.
Dirk
Travis Dixon
sparcboy
5500. We tried setting the Cisco to auto-negotiate with no luck, we fixed
both ends at 100 full duplex (use ndd on Solaris) and have never had another
problem.
sparcboy
>===== Original Message From Dirk Hufnagel <hufn...@mps.ohio-state.edu> =====
Paul Lavoie
Travis Dixon <tra...@shell.clark.net> wrote:
>dwm...@my-deja.com wrote:
>>> Buy better switches or upgrade the firmware in the switches.
>> Or you could switch to HP 9000 :>
>
>...and get the exact same problems. Sorry, I've had to
>do the same thing on my HP's as I've done to the suns --
>lock the speed and duplex. Nobody seems to play well
>with Cisco switches...
Anybody know how to deal with 2 interfaces where one wants to be
10 half and the other 100 full?. The NDD commands do not seem to support
more than one instance of the interface. Both interfaces are hme's.
--
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Paul R Lavoie Paul....@Software.com
Network Services Manager Office:(781) 674-0100 x275
Software.com. Cell:(617) 901-1151
10 Maguire Road, Lexington Ma 02421 Fax:(781) 674-1080
Software.com, Inc. The Internet Infrastructure Company (tm)
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
Peter Schmid
Paul Lavoie wrote:
> In article <h2Nj4.55509$Rj5....@dfw-read.news.verio.net>,
> Travis Dixon <tra...@shell.clark.net> wrote:
> >dwm...@my-deja.com wrote:
> >>> Buy better switches or upgrade the firmware in the switches.
> >> Or you could switch to HP 9000 :>
> >
> >...and get the exact same problems. Sorry, I've had to
> >do the same thing on my HP's as I've done to the suns --
> >lock the speed and duplex. Nobody seems to play well
> >with Cisco switches...
>
> Anybody know how to deal with 2 interfaces where one wants to be
> 10 half and the other 100 full?. The NDD commands do not seem to support
> more than one instance of the interface. Both interfaces are hme's.
You can change which device you are working with by using:
ndd -set /dev/hme instance 0 or 1 or whichever number the device is. Then
you can change/view the parameters
Pete.
Lon Stowell
Paul Lavoie <lav...@pilsner.software.com> wrote:
>In article <h2Nj4.55509$Rj5....@dfw-read.news.verio.net>,
>Travis Dixon <tra...@shell.clark.net> wrote:
>>dwm...@my-deja.com wrote:
>>>> Buy better switches or upgrade the firmware in the switches.
>>> Or you could switch to HP 9000 :>
>>
>>...and get the exact same problems. Sorry, I've had to
>>do the same thing on my HP's as I've done to the suns --
>>lock the speed and duplex. Nobody seems to play well
>>with Cisco switches...
>
> Anybody know how to deal with 2 interfaces where one wants to be
> 10 half and the other 100 full?. The NDD commands do not seem to support
> more than one instance of the interface. Both interfaces are hme's.
From the command line:
ndd -set /dev/hme instance 0
muck with 0
ndd -set /dev/hme instance 1
muck with hme1
tony_b...@my-deja.com
c...@NOSPAM.netunix.com (C. Newport) wrote:
> Jonathan Magid (j...@triphop.metalab.unc.edu) wrote:
> : In article <3882A45D...@dimensional.com>,
> : Robert Yoder <ryo...@dimensional.com> wrote:
> : >
> : >If your network device is named "hme*" or "qfe*"
> : >it is a 10/100 device.
> : [suggestions deleted]
>
> : which are running at half duplex and have be* ethernet cards. The
> : UE4000 that's on the same switch correctly detects the network's
> : capability for full duplex communications.
>
>
running
> : full or half duplex under Solaris 2.7?
grep hme0 /var/adm/messages | grep -i duplex
>
> Solaris 7 does not support the be device.
> ( according to the release notes ).
>
>
Heath Muchmore
we type the following from the command line as per sunsolve.sun.com faq 2605.
did this so we didn't have to reboot. and we did this for hme and qfe.
ndd -set /dev/hme instance 0
ndd -set /dev/hme adv_100fdx_cap 1
ndd -set /dev/hme adv_100hdx_cap 0
ndd -set /dev/hme adv_10fdx_cap 0
ndd -set /dev/hme adv_10hdx_cap 0
ndd -set /dev/hme adv_autoneg_cap 0
Then we added the following to /etc/system for reboots.
set hme:hme_adv_autoneg_cap=0
set hme:hme_adv_100T4_cap=0
set hme:hme_adv_100fdx_cap=1
set hme:hme_adv_100hdx_cap=0
set hme:hme_adv_10fdx_cap=0
set hme:hme_adv_10hdx_cap=0