Comware

I have a site that has a HP8000M switch acting as a core switch with fiber uplinks to 5 HP 4000M switches in which some are trunked with 2 fiber ports to the other HP 4000M switches. There are also 4 ProCurve 2524 switches uplinked to the HP 8000M switch via Fiber modules with no trunking.

On all of the 2524 switches flow control is disabled on all ports. On the HP8000M switch all ports are set with flow control enabled. On some of the HP4000M switches some of the ports are set to flow control enabled and some disabled but most have the trunk ports setup to flow control enabled. I thought that it was recommended to have all user ports set to flow control enabled and all other ports set to flow control disabled on HP switches. What is the standard recommendation on HP switches for setting flow control.

They are having a problem with a few times a day the fiber trunk ports to one of the HP 4000m switch from the HP8000M switch locks up and all network traffic stops flowing through the HP 8000M. As soon as the ports lock between these two switches you see the other fiber connections one by one lock up. When I say lock up the activity LEDs on the fiber ports go solid. You cannot ping any device across the network. They usually can unplug the two fiber connections between the two fiber ports to the intial HP 4000M switch that has started the problem and then connect them back up after about 60 seconds and the network will typically come back. Sometimes however they have had to physically reset the HP 8000M and HP4000M switches that initially had the problem.

Am assuming that they are having a device on the network creating excessive traffic when this happens and will be onsite tomorrow to run a sniffer trace but am curious about any suggestions on flow control and if others have seen this problem.

Thanks for any information!

Brad,
1. Is the firmware of the switches updated to the latest version? I remember I had some problems with older versions.
2. What kind of fiber modules you use? 100 Mbps, 1Gbps?

All of the switches have been upgraded to the latest firmware over the weekend.

In the HP8000M there are 3 4-port 100Mbps cards and there is 1 1-port 1Gbps card. The 4000M switches have 1 4-port 100Mbps cards in them.

Hello Brad,

with this hardware setup you are driving the 8000M at the limits. or basically you make it possible to run into congestion.

basically this makes flow control "necessarry" on the core devices weather the recommendation is for the edge devices or not. your edge devices "could use flow control" to decide which traffic needs to be sent through the 100mbit uplink. so they are the bottlenecks.

regarding this setup I basically would consider to switch the whole setup to "no flow control" and when the traffic changed to a level that does not work anymore upgrade the core device and use gigaabit uplinks. that is safe.

the locking of your devices bassically pretty much sounds not like a flow control error but some amok stp or broken trunks.