Gigabit Interface, losing connection.

[Cristian_4]

Hi,

I got a gigabit interface in a hpux 11.0, it works ok but after a transfer 300k+ (ftp) the interface stop responding. The only workaround is to reboot the server, an ifconfig down and up doesnt work.

any suggestion ?

thanks

[Tony Scully_2]

Christian,

What model card is it, what model server and what patch revision are you at?

Post ioscan -fnuClan, swlist, ifconfig .

Cheers,

Tony

[Cristian_4]

Hi

The server is a j5600

# ioscan -fnuClan
Class I H/W Path Driver S/W State H/W Type Description
====================================================================
lan 0 10/0/12/0 btlan3 CLAIMED INTERFACE HP PCI 10/100Base-TX Core
/dev/diag/lan0 /dev/ether0
lan 1 10/1/5/0 gelan CLAIMED INTERFACE HP A4929A PCI 1000Base-T Adapter
/dev/gelan1
# swlist
# Initializing...
# Contacting target "host"...
#
# Target: host:/
#

#
# Bundle(s):
#

64-bitDevLibs B.11.00 Cross Platform Development Kit
A3495A B.11.00.06 HP-PB 100BT LAN/9000
A3722A B.11.00.07 HSC FDDI/9000 Driver for K,T Class
A3723A B.11.00.07 HSC FDDI/9000 Driver for B,C,D,J Class
A3739B B.11.00.16 PCI FDDI/9000 Media
A4919A B.11.00.12 HyperFabric 9000/[78]00
A4924A B.11.00.11 1000Base-SX HSC Gigabit Ethernet Driver
A4926A B.11.00.11 1000Base-SX PCI Gigabit Ethernet Driver
A4929A B.11.00.11 1000Base-T PCI Gigabit Ethernet Driver
A5158A B.11.00.10 PCI Tachyon TL/TS Fibre Channel
A5230A B.11.00.05 100BT/9000 PCI
A5506A B.11.00.02 PCI 4 PORT 100BASE-T/9000
A5506B B.11.00.02 PCI 4 PORT 100BASE-T/9000
A5783A B.11.00.07 PCI Token Ring
A5856A B.11.00.04 I2O RAID
A6092A B.11.00.12 HyperFabric 9000/[78]00
A6684A B.11.00.10 HSC Fibre Channel
A6685A B.11.00.10 HSC Fibre Channel
A6748A B.11.00.11 PCI RS-232 MUX Software
A6795A B.11.00.10 PCI Tachyon TL/TS/XL2 Fibre Channel
B5509BA B.11.00.05 100BT/9000 PCI
HPUXEng64RT B.11.00.01 English HP-UX 64-bit Runtime Environment
HWE1100 B.11.00.0203.5 Hardware Enablement Patches for HP-UX 11.00, March 2002
Integ-Logon B.11.00 Integrated Logon Bundle
J2157B B.11.00.08 Server FDDI/9000
J2166B B.11.00.07 Token Ring/9000 HPPB Series 800
J3515A B.11.00.05 100BT/9000 GSC for B, C, D, and J Series
J3516A B.11.00.05 100BT/9000 GSC for B, C, D, and J Series
KernDev B.11.00 HPUX Kernel Developers Kit
OnlineDiag B.11.00.20.09 HPUX 11.0 Support Tools Bundle, Mar 2002
QPK1100 B.11.00.56.5 Quality Pack for HP-UX 11.00, March 2002
UXCoreMedia B.11.00.02 HP-UX Media Kit (Reference Only. See Description)
UnlimUserLic B.11.00.02 HP-UX Unlimited-User License
XSWGR1100 B.11.00.47.08 General Release Patches, November 1999 (ACE)
# ifconfig lan1
lan1: flags=1842
inet 192.168.0.45 netmask ffffff00 broadcast 192.168.0.255
#

Lan0 is down btw

Thanks

[Tony Scully_2]

Christian,

Is this a new card, or has this behaviour just appeared? Anything changed (switch side maybe?) recently?

When the card 'hangs' I assume the ifconfig looks fine, no errors logged anywhere?

Do you have a spare card you could swap this out for?

Sorry to be of little help.

Tony.

[Cristian_4]

it's a new card and i dont have a spare =(

nothing has changed, there arent any logs and ifconfig is fine when it happens, the card seems to be ok, in fact i got green lights all the time.

thanks

[Tony Scully_2]

Christian,

Could try running stm (I find xstm easier) and run the exercise option on the card, this might at least give some sort of error.

Tony

[Cristian_4]

i did the test and i do get errors, but there's nothing about the possible error.

[Rita C Workman]

...just a silly question...

are you using cat 5 cable ? or cat 6 ?

/rita

[Cristian_4]

cat 5, but it works with a linux server

[Tony Scully_2]

Christian,

Just checking - the problem happens with any protocol after a time, not just ftp?

Is it possible to eliminate the rest of the network, ie attach to the card with a crossover cable and ftp over that.

Clutching at straws I'm afraid.
Tony

[Jim Keeble]

If it's not installed already, You might try the current gelan patch, PHNE_28995. This raises the driver version from 11.00.11 to 11.00.21. There are a couple of "hang" problems fixed.

Also, rather than rebooting, you might try a card reset from within the lanadmin menu's. This is not a fix, but at least a datapoint that might be useful if you need to open a support case.

If the node is a vpar (Vitual Partition) and the vpar version is old (2.02 or 2.03 I think) the curent VPAR version 3.02 will help with odd I/O problems.

[Cristian_4]

its already installed, this happens with every protocol

the max. amount of transfer is always 128k

[rick jones]

IIRC, Gigabit wants Cat5e or better. That the cable "works" with another server is good, but not conclusive since that other server is probably using a different MAC and in particular PHY. Perhaps with different tolerances.

Is the problem seen after a transfer of _any_ file, or does it happen with the transfer of a particular file? If a particular file, if you gzip the file (or ungzip it) does the behaviour change?

Can you reproduce the failure with say a netperf TCP_STREAM test rather than an FTP transfer?

When the link appears to be down, can you try a linkloop command instead of ping? This tends to require another HP-UX system because others may not respond to the XID/TEST frame that linkloop (think link-level ping) does.

Any chance another system may have the same IP?

The suggestion to try a lanadmin reset is a very good one.

An A4929A is not exactly a "new" GbE NIC in that the design goes back a ways - it is based on an Alteon Tigon2 NIC. But, I'm not sure that the J6825A was ever qualified on the J5600.

Might make sure that the latest PCI and/or I/O patches are installed as well.