Procurve 4000m off-line/on-line port

[noel_28]

hi
in the log of the switch I see for the ports a off/on during 2 seconds.

---------------------------------------

I 02/17/05 05:38:24 ports: port E6 is now off-line
I 02/17/05 05:38:26 ports: port E6 is now on-line
I 02/17/05 05:59:48 ports: port E7 is now off-line
I 02/17/05 05:59:51 ports: port E7 is now on-line
I 02/17/05 06:00:17 ports: port E7 is now off-line
I 02/17/05 06:00:20 ports: port E7 is now on-line
I 02/17/05 07:26:35 ports: port D3 is now off-line
I 02/17/05 07:26:37 ports: port D3 is now on-line
I 02/17/05 07:27:05 ports: port D3 is now off-line
I 02/17/05 07:27:07 ports: port D3 is now on-line
I 02/17/05 07:47:08 ports: port C8 is now off-line
I 02/17/05 07:47:10 ports: port C8 is now on-line

we diffuse the video one via multicast and flow is cut when there is a off/on.
the firmware is C.09.22 the configuration is simple, in attachement the file of configuration.

how to remove these off/on?

Best regard's

[Les Ligetfalvy]

It may simply be stating the truth, that the link is toggling off/on. Take a look at the power management setting for the NIC.

[Ron Kinner]

Did you check for errors on the ports? Some switches will cycle a port if they get too many errors. Usually this is caused by a DUPLEX mismatch. Somebody sets all of the switch ports for FULL when the hosts are all set to Auto or vice versa. This will always cause a duplex mismatch with excessive FCS and Runt errors on the FULL side and excessive collisions on the Auto side which will, of course, be worse during a high traffic situation like a multicast. Both sides must be set to FULL or both sides set to Auto.

Ron

[noel_28]

the two sides are with AUTO (switch and PC). There are no errors on the ports and wiring is new and is certified.
The PC have either of INTEL pro 1000 card under linux Debian 3.0 with the last driver of INTEL or of the Realteck card under Windows 2000 Pro SP4 with the last driver of the manufacturer.

i try to set FULL duplex on the two sides.

Thanks

[Anton Pikhteryev_2]

I have exactly the same problem:

I 10/26/05 09:06:43 ports: port I4 is now off-line
I 10/26/05 09:06:46 ports: port I4 is now on-line
I 10/26/05 09:32:35 ports: port B2 is now off-line
I 10/26/05 09:32:39 ports: port B2 is now on-line
I 10/26/05 09:32:47 ports: port B2 is now off-line
I 10/26/05 09:32:49 ports: port B2 is now on-line
I 10/26/05 09:33:11 ports: port B2 is now off-line
I 10/26/05 09:33:13 ports: port B2 is now on-line
I 10/26/05 09:34:22 ports: port I5 is now off-line
I 10/26/05 09:34:22 ports: port I6 is now off-line
I 10/26/05 09:34:25 ports: port I5 is now on-line
I 10/26/05 09:34:25 ports: port I6 is now on-line

None of the devices connected to the listed ports went off-line. Uptime on these devices didn't change. It is something wrong with the HP switch.

Firmware revision : C.09.22
ROM Version : C.06.01

I would like to get this issue resolved ASAP, any comments on it?

Thank you

[OLARU Dan]

I think you should enable IGMP for the VLAN you use multicast on in this switch menu: "Switch Configuration - Advanced Features - IGMP Service"

[Anton Pikhteryev_2]

how does IGMP relate to the directly connected link goes offline/online?

[Michael Enk]

I am seeing similar problems with many ports going off-line then on-line on several 5372XL switches. IGMP and STP are disabled. We have not been able to find the cause of this. If anyone has a solution I would appriciate the help.

Best regards,
Michael

[VinceWhirlwind]

Same here - been having this problem for the past 3/4 months. When it's bad it causes the network to underperform very significantly.

We have recently been adding Cisco switches to our network.

I have 4108, 4000, 2650, 2524 HP switches.

I'v been told that *EN*abling IGMP is good - the switch learns IGMP paths which improves performance.

I found the problem was solved on the non-4000 switches when I disabled LACP.

So, the question is:
1/ Do the 4000s use LACP?
2/ How to disable LACP on the 4000s?

On the other switches, the command is :
no interface all lacp
but the 4000 uses set commands and I can't find one that is relevant....

[Michael Enk]

I have LACP disabled on all interfaces on several 5372XL's without any improvements. I have users, who are thrown off the network every 10 minutes now. The strange part is that it happened just 3 days ago without any warning even a reboot of the switches solved nothing. It sort of tells me that something was introduced into our network, which causes this e.g. by confusing the switches mac/arp tables, however I have not been able to figure out what this something might been, which could impact a network like this.