Problems with powersave?

[sunshinedays]

Are there any known issues with procurve switches and PC's that are in suspend mode. We are seeing something along the lines of:-

PC (for example dell optiplex 360) goes into sleep mode.The switch (eg 4202vl) port shows active but drops to 10MB and show mac-address shows no mac address. The switch shows various alerts about high collision or drop rate, presumably from broadcast traffic.

A related issue is that when said machines come back on the switch does not always change the speed back despite everything being set to auto and we get lots (millions) of transmit drops.

The first issue happens on all our switches while the second one is just on one that I've found. The network is inherited and is a mix of various procurves,. Wer’e also getting lots general network issues and suspect, but cannot find either a loop or some strange duplex related goings on.

[Matcol]

The first thing is to map out the network: log onto each switch and do a "show lldp inf rem". Make a diagram of how everything is interconnected and label each device's IP address.

The next thing is to get an SNMP string onto all of them and then get some sort of network monitoring on it, so you can see any ports showing high utilisation or error rate.

Then, setup a syslog server and point all your network equipment at it.

Then, while your stats are being collected make a catalogue of all your different switch models, then get on the HP website and download the latest firmware for each model. Then upgrade all switches so they all have the same - latest - firmware per model.

[dmesser-hhs]

I am experinceing similar issues with all of my procurve switches.  This problem just started happening for me.  I was getting ready to implement spanning tree and I started getting the 00331 FFI high collision or drop rate messages on all of my switches.  These messages come in on about 4 or 5 ports on each switch about once an hour.

I have done as suggested I did a portmap on all my switches (I did this prior to implementing spanning tree)  one thing I did prior to turning on spanning tree was to turn on loop protect.  I plan on going back and turning off loop protect on all 50 switches.

My switch models range from 2910, 3400, 3500, 5412, 8212.

This is only happening at my main campus.  I have another campus that runs 5400 and 2900 only and I have not gotten one of these messages there.

[sunshinedays]

Not sure if this helps but I did eventually find the cause of the original issues. Essentially it was microsoft network load balancer software not playing well with our switch. It seems there are known problems with it and some switches that whoever first installed it was not aware of.

In our case the traffic flow went along the lines of

1. Request sent to our sharepoint server using the load balancer

2. The load balancer was a virtual machine set to use multicast traffic and sent multicast data to the web front ends

3.HP switch model did not fully undertand the traffic type so sent the requests everywhere.

4. Since the traffic went everywhere, including the servers it was intended for it appeared to work.

The problem had been there for years but it was only the increase in use that made it apparent so if you have a similar setup it could be a spike in normal traffic going to the wrong places.

[dmesser-hhs]

Thank you so much for the post.  Yes I have a MS load-balance application running.  And when I did a packet capture I noticed the packets from this server.  As a matter of fact if I do a "show top talker" for any port on my switches the load balancer always shows up in the top spot.

So how did you mitigate the problem, or did you just live with it?  Do I need to turn on multi-cast function for my switches?

[sunshinedays]

I don't advise trying to live with it. For us It got a lot worse before we cracked it and it was causing major issues during an OFSTED inspection so we pulled the second server and bypassed the load balancer. The server has been running ok without it and we've since replaced all our switches as part of a move to a new building.

The firmware on the core switch was found to be a bit old, as were many of the edge switches so it might be that a firmware upgrade might help. If not already, having your servers in their own vlan might reduce the issue.

From what I read the load balancer uses somewhat non standard mac addresses which was the problem rather than the switche's ability to handle muticast traffic so you might be better off looking at the load balancer. Apparently the software can be set to not use multicatst but I thinkit would need to not be a virtual machine. Here is a good description of a similar issue I found once I knew what to be looking for:-

http://community.kaseya.com/xsp/f/114/t/13653.aspx