I have a pkt-loss issue when running Speed Flex on access points at certain building locations. We have a ZD3000 at our main site, and the access points are behaving as normal. At our two other sites, each connected by a 10mb fast Ethernet connection, the access points report a ~15% downlink pkt-loss and a ~50% for one site to ~80% for the other site uplink pkt-loss. They did not always have this issue. We recently placed HP e5406zl switches at each end of these fast Ethernet connections replacing routers. If the problem occurred at the time of the switch over, I did not notice it, but we had been running on them for a good month. Last week, I had a VLAN added to each switch so that I could plug in my PoE IP phones. I believe this marked the time when the problems started.
The access points are not plugged into this new VLAN. The access points are getting their correct DHCP addresses after a power cycle. (The new VLAN serves DHCP addresses from a completely different class c range.) When we made the routing changes on the HP switches, we did not alter existing working data routes, only added phone related routes that shunt traffic to dedicated T1 lines for phones. We have since confirmed that data traffic directed to either end's HP switch (which is also the gateway address for each location) passes the traffic back and forth over the fast Ethernet.
I'm at a loss as to how to continue troubleshooting the issue. Where else should I be looking to address this pkt-loss issue?
I did correct a subnet issue on our ZD3000 where I had made the scope too generous (255.255.240.0 instead of 255.255.248.0). It appeared that broadcast transmissions would not be routed by the HP switches across the fast Ethernet from the ZD3000, where they previously had been with the routers. This fixed a 0% downlink report from the remote site access points, but didn't impact the pkt-loss issue.
Thank you for any assistance that you may provide.
I assume that everything else works, AP are connected to the ZD, clients can connect and pass traffic on the correct VLAN and so on and that the only error is the packet drop when running speedflex, correct?
This can be a side effect of how speedflex works. When you run speedflex, you create two endpoint that will be exchanging traffic and reporting the results to another system. If you run speedflex between a ZD and an AP and both are connected to the same subnet and they both are on gig connection, you will not see any packet loss, but if the AP is connected to a 100 Mbps instead of a gig like the ZD then the downlink (ZD to AP) will experience packet loss where the uplink will not. This is because when you send UDP traffic form the ZD you are sending it at a gig rate and the AP is only able to receive at 100 Mbps. Not all the packet at a rate of 600 Mbps from the ZD will be able to be recieved by the AP, so you will see packet loss on the downlink. When the donwlink is done the AP will then start send traffic back to the ZD, but since it's connected at 100 Mbps it will only send packet at about 90 Mbps which can be easily received at the ZD's gig connection easily so you do not see packet loss.
You might consider using ZAP instead of speedflex and running at a rate at the lowest common denominator, -r 10 for example since you have a 10 Mbps link between the two testing points and use the TCP (-t) option instead of the default UDP.