I have a strange problem here that maybe someone here could shed some
light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2 ML1 on a
couple of 6h1 servers. Both of these server have two Gigbit SL PCI
NIC's setup in as an etherchannel. Communications between these two
server (both connected to the same switch) bombs when a large packet
goes through. I've yet to narrow it down to how big of a packet...
but I noticed that I can ftp files of less than 1K but can't for files
much more than that. A zero byte file gets created on the recieving
machine and the the ftp process hangs SFTP, large db queries and such
all hang. I CAN however transmit large files between other machines
and these systems. This includes another 6h1 with a baseT ethernet
card which also was upgraded to exactly the same level and connected
to the same switch as the two problem machines.

Is there something maybe with the MTU or send/recieve sizes that
change during the upgrade? Does it sound like an etherchannel issue?
Do the network parameters need to be changed? Any ideas would be much
appreciated.

thanks,

-Patrick

Are these Cisco switches ? We had the same problem, and Cisco had a patch for some problem on the switch
and it fixed this.

Dennis


"danko" <dankopat@washburn.edu> wrote in message news:983dfde8.0311251005.1d55b3e5@posting.google.c om...
> I have a strange problem here that maybe someone here could shed some
> light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2 ML1 on a
> couple of 6h1 servers. Both of these server have two Gigbit SL PCI
> NIC's setup in as an etherchannel. Communications between these two
> server (both connected to the same switch) bombs when a large packet
> goes through. I've yet to narrow it down to how big of a packet...
> but I noticed that I can ftp files of less than 1K but can't for files
> much more than that. A zero byte file gets created on the recieving
> machine and the the ftp process hangs SFTP, large db queries and such
> all hang. I CAN however transmit large files between other machines
> and these systems. This includes another 6h1 with a baseT ethernet
> card which also was upgraded to exactly the same level and connected
> to the same switch as the two problem machines.
>
> Is there something maybe with the MTU or send/recieve sizes that
> change during the upgrade? Does it sound like an etherchannel issue?
> Do the network parameters need to be changed? Any ideas would be much
> appreciated.
>
> thanks,
>
> -Patrick

In article <lTexb.5832$aw2.2376934@newssrv26.news.prodigy.com >, News Reader
<NewsReader@nowhere1else.con> wrote:
> Are these Cisco switches ? We had the same problem, and Cisco had a
> patch for some problem on the switch and it fixed this.

What switch model was it, and what software revision fixed it, if I could
ask? We're significantly Cisco here so this is useful stuff.

-Dan

Dan Foster <dsf@globalcrossing.net> wrote in message news:<slrnbsavf0.r9d.dsf@gaia.roc2.gblx.net>...
> In article <lTexb.5832$aw2.2376934@newssrv26.news.prodigy.com >, News Reader
> <NewsReader@nowhere1else.con> wrote:
> > Are these Cisco switches ? We had the same problem, and Cisco had a
> > patch for some problem on the switch and it fixed this.
>
> What switch model was it, and what software revision fixed it, if I could
> ask? We're significantly Cisco here so this is useful stuff.
>
> -Dan

We are an all Cisco shop here. I'm not sure the model switch off the
top of my head but I'll check Monday. Could you post the patch info
please? If that takes care of it you've got a free drink coming to
you on me.

Thanks,

-Patrick

"News Reader" <NewsReader@nowhere1else.con> wrote in message news:<lTexb.5832$aw2.2376934@newssrv26.news.prodig y.com>...
> Are these Cisco switches ? We had the same problem, and Cisco had a
> patch for some problem on the switch
> and it fixed this.
>
> Dennis
>
>
> "danko" <dankopat@washburn.edu> wrote in message
> news:983dfde8.0311251005.1d55b3e5@posting.google.c om...
> > I have a strange problem here that maybe someone here could shed some
> > light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2 ML1 on a
> > couple of 6h1 servers. Both of these server have two Gigbit SL PCI
> > NIC's setup in as an etherchannel. Communications between these two
> > server (both connected to the same switch) bombs when a large packet
> > goes through. I've yet to narrow it down to how big of a packet...
> > but I noticed that I can ftp files of less than 1K but can't for files
> > much more than that. A zero byte file gets created on the recieving
> > machine and the the ftp process hangs SFTP, large db queries and such
> > all hang. I CAN however transmit large files between other machines
> > and these systems. This includes another 6h1 with a baseT ethernet
> > card which also was upgraded to exactly the same level and connected
> > to the same switch as the two problem machines.
> >
> > Is there something maybe with the MTU or send/recieve sizes that
> > change during the upgrade? Does it sound like an etherchannel issue?
> > Do the network parameters need to be changed? Any ideas would be much
> > appreciated.
> >
> > thanks,
> >
> > -Patrick


The switch these two servers are going into is a 3508 on 12.0(5.2)XU.
Again, could you post details of the patch you applied?

thanks much,

-Patrick

"danko" <dankopat@washburn.edu> wrote in message news:983dfde8.0312030628.4fd8277f@posting.google.c om...
> "News Reader" <NewsReader@nowhere1else.con> wrote in message news:<lTexb.5832$aw2.2376934@newssrv26.news.prodig y.com>...
> > Are these Cisco switches ? We had the same problem, and Cisco had a
> > patch for some problem on the switch
> > and it fixed this.
> >
> > Dennis
> >
> >
> > "danko" <dankopat@washburn.edu> wrote in message
> > news:983dfde8.0311251005.1d55b3e5@posting.google.c om...
> > > I have a strange problem here that maybe someone here could shed some
> > > light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2 ML1 on a
> > > couple of 6h1 servers. Both of these server have two Gigbit SL PCI
> > > NIC's setup in as an etherchannel. Communications between these two
> > > server (both connected to the same switch) bombs when a large packet
> > > goes through. I've yet to narrow it down to how big of a packet...
> > > but I noticed that I can ftp files of less than 1K but can't for files
> > > much more than that. A zero byte file gets created on the recieving
> > > machine and the the ftp process hangs SFTP, large db queries and such
> > > all hang. I CAN however transmit large files between other machines
> > > and these systems. This includes another 6h1 with a baseT ethernet
> > > card which also was upgraded to exactly the same level and connected
> > > to the same switch as the two problem machines.
> > >
> > > Is there something maybe with the MTU or send/recieve sizes that
> > > change during the upgrade? Does it sound like an etherchannel issue?
> > > Do the network parameters need to be changed? Any ideas would be much
> > > appreciated.
> > >
> > > thanks,
> > >
> > > -Patrick
>
>
> The switch these two servers are going into is a 3508 on 12.0(5.2)XU.
> Again, could you post details of the patch you applied?
>
> thanks much,
>
> -Patrick

Since I'm not in the network group, I don't know the answers to your questions. I sent a
query to one of the network guys but have not got a response yet. The only thing I know for
sure is that they were new Cisco switches, and they had to apply some patch to fix the problem.
The symptoms on my end were identical to what you describe. I placed a 4 port hub between
my lpars, and the problem went away. That is how I convinced the network group the problem
was on their end.

Dennis

"News Reader" <NewsReader@nowhere1else.con> wrote in message news:<_2Szb.9082$aw2.4264944@newssrv26.news.prodig y.com>...
> "danko" <dankopat@washburn.edu> wrote in message
> news:983dfde8.0312030628.4fd8277f@posting.google.c om...
> > "News Reader" <NewsReader@nowhere1else.con> wrote in message
> news:<lTexb.5832$aw2.2376934@newssrv26.news.prodig y.com>...
> > > Are these Cisco switches ? We had the same problem, and Cisco had
> a
> > > patch for some problem on the switch
> > > and it fixed this.
> > >
> > > Dennis
> > >
> > >
> > > "danko" <dankopat@washburn.edu> wrote in message
> > > news:983dfde8.0311251005.1d55b3e5@posting.google.c om...
> > > > I have a strange problem here that maybe someone here could shed
> some
> > > > light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2 ML1
> on a
> > > > couple of 6h1 servers. Both of these server have two Gigbit SL
> PCI
> > > > NIC's setup in as an etherchannel. Communications between these
> two
> > > > server (both connected to the same switch) bombs when a large
> packet
> > > > goes through. I've yet to narrow it down to how big of a
> packet...
> > > > but I noticed that I can ftp files of less than 1K but can't for
> files
> > > > much more than that. A zero byte file gets created on the
> recieving
> > > > machine and the the ftp process hangs SFTP, large db queries and
> such
> > > > all hang. I CAN however transmit large files between other
> machines
> > > > and these systems. This includes another 6h1 with a baseT
> ethernet
> > > > card which also was upgraded to exactly the same level and
> connected
> > > > to the same switch as the two problem machines.
> > > >
> > > > Is there something maybe with the MTU or send/recieve sizes that
> > > > change during the upgrade? Does it sound like an etherchannel
> issue?
> > > > Do the network parameters need to be changed? Any ideas would be
> much
> > > > appreciated.
> > > >
> > > > thanks,
> > > >
> > > > -Patrick
> >
> >
> > The switch these two servers are going into is a 3508 on 12.0(5.2)XU.
> > Again, could you post details of the patch you applied?
> >
> > thanks much,
> >
> > -Patrick
>
> Since I'm not in the network group, I don't know the answers to your
> questions. I sent a
> query to one of the network guys but have not got a response yet. The
> only thing I know for
> sure is that they were new Cisco switches, and they had to apply some
> patch to fix the problem.
> The symptoms on my end were identical to what you describe. I placed a
> 4 port hub between
> my lpars, and the problem went away. That is how I convinced the
> network group the problem
> was on their end.
>
> Dennis


Thanks for the reply. I appreciate any follow up you can give me from
your network group.

thanks again,

-Patrick

> > The symptoms on my end were identical to what you describe. I placed a
> > 4 port hub between
> > my lpars, and the problem went away. That is how I convinced the
> > network group the problem
> > was on their end.
> >
> > Dennis
>
>
> Thanks for the reply. I appreciate any follow up you can give me from
> your network group.
>
> thanks again,
>
> -Patrick

To ascertain where the problem lies, have you followed the suggestion
above of putting a hub (instead of a switch) in place and seeing what
the results are ?

Steve

steve@wakefieldrfc.freeserve.co.uk (Steve Nottingham) wrote in message news:<42862645.0312240417.651907b9@posting.google. com>...
> > > The symptoms on my end were identical to what you describe. I placed a
> > > 4 port hub between
> > > my lpars, and the problem went away. That is how I convinced the
> > > network group the problem
> > > was on their end.
> > >
> > > Dennis
> >
> >
> > Thanks for the reply. I appreciate any follow up you can give me from
> > your network group.
> >
> > thanks again,
> >
> > -Patrick
>
> To ascertain where the problem lies, have you followed the suggestion
> above of putting a hub (instead of a switch) in place and seeing what
> the results are ?
>
> Steve


Unfortunately the NICs on these two machines are fiber. The only
network equipment we have lying around that has fiber capability are
switches and routers (all cisco).

-Patrick

danko <dankopat@washburn.edu> wrote:
> I have a strange problem here that maybe someone here could shed
> some light on (I hope!). Recently upgraded from 4.3.3 ML11 to 5.2
> ML1 on a couple of 6h1 servers. Both of these server have two
> Gigbit SL PCI NIC's setup in as an etherchannel. Communications
> between these two server (both connected to the same switch) bombs
> when a large packet goes through. I've yet to narrow it down to how
> big of a packet... but I noticed that I can ftp files of less than
> 1K but can't for files much more than that. A zero byte file gets
> created on the recieving machine and the the ftp process hangs SFTP,
> large db queries and such all hang. I CAN however transmit large
> files between other machines and these systems. This includes
> another 6h1 with a baseT ethernet card which also was upgraded to
> exactly the same level and connected to the same switch as the two
> problem machines.

> Is there something maybe with the MTU or send/recieve sizes that
> change during the upgrade? Does it sound like an etherchannel issue?
> Do the network parameters need to be changed? Any ideas would be
> much appreciated.

I cannot speak to AIX specifically, but if the two recently upgraded,
Gigabit connected systems cannot talk to one another with "large"
packets, perhaps they have inadvertently enabled jumbo frame support
in the NICs?

IIRC, the Cisco switches do not unversally support Jumbo Frames.

If you can take a packet trace (tcpdump,snoop, whatever AIX supports),
you might look at the MSS negotiation for a connection between the two
systems, and compare that with the MSS negotiation with the "plain"
ethernet connected systems.

Another way to more blindly try that out would be to take the file
sizes up to a typical ethernet mss of 1460 bytes and then try 1461.
Packet trace would be better though...

rick jones
--
portable adj, code that compiles under more than one compiler
these opinions are mine, all mine; HP might not want them anyway...
feel free to post, OR email to raj in cup.hp.com but NOT BOTH...