Unsolved
This post is more than 5 years old
16 Posts
0
6333
January 24th, 2007 11:00
Dell PowerConnect 6024F does not respect it’s ARP table
We have a Dell PowerConnect 6024F switch that is used as a CORE switch with several interfaces / VLANs. The switch is running with the latest firmware - v.2.0.0.13 , A03. All interfaces, except 2, is configured as /25 (255.255.255.128) with 10.x.x.x IPs.
Everything was running fine until we added a new interface for VoIP usage. This interface is configured as /23 (255.255.254.0). Now we experience a periodic (everything from a few hours to many days constant) overflow of ARP packets on this single interface. ARP traffic on the other interfaces is normal.
Whenever the IP Phone (from Nortel) want's to communicate with the server (or vice versa), the Dell switch broadcasts ARP to learn the devices MAC address. But if I look in the switch's ARP table, all the devices it broadcasts ARP to learn the MAC address, is already available in the ARP table, but it seems like the switch ignores it. I've tested this by sniffing the ARP traffic with Ethereal and comparing with the switch's ARP table.
All this ARP traffic seems to "kill" the VoIP interface, because if I try to ping the VoIP server while we experience the problem, I get a 70% packet loss and the IP phones are therefore impossible to use. I've tried graphing the traffic on the interface with MRTG, and it shows a constant load of approx. 10KB/s of ARP traffic and Ethereal sniffed more than 20.000 ARP packets in a 6-minute period - that's more than 50 ARP packets per second on an interface with less than 150 devices!
We've tried restarting the switch (both cold and warm restart) and clearing it's ARP table manually, but nothing helped. ARP timeout is set to 60.000 seconds so it's not because the switch purges the devices MAC address.
We haven't implemented QoS on the network (all Dell PowerConnect switches), yet.
Dell Business Support has of course been contacted (service level is 4 hours Gold), but they would not help us because they currently don't support VoIP on Dell switches!
Any advice is much appreciated!
Thanks in advance.
/MiGe
Everything was running fine until we added a new interface for VoIP usage. This interface is configured as /23 (255.255.254.0). Now we experience a periodic (everything from a few hours to many days constant) overflow of ARP packets on this single interface. ARP traffic on the other interfaces is normal.
Whenever the IP Phone (from Nortel) want's to communicate with the server (or vice versa), the Dell switch broadcasts ARP to learn the devices MAC address. But if I look in the switch's ARP table, all the devices it broadcasts ARP to learn the MAC address, is already available in the ARP table, but it seems like the switch ignores it. I've tested this by sniffing the ARP traffic with Ethereal and comparing with the switch's ARP table.
All this ARP traffic seems to "kill" the VoIP interface, because if I try to ping the VoIP server while we experience the problem, I get a 70% packet loss and the IP phones are therefore impossible to use. I've tried graphing the traffic on the interface with MRTG, and it shows a constant load of approx. 10KB/s of ARP traffic and Ethereal sniffed more than 20.000 ARP packets in a 6-minute period - that's more than 50 ARP packets per second on an interface with less than 150 devices!
We've tried restarting the switch (both cold and warm restart) and clearing it's ARP table manually, but nothing helped. ARP timeout is set to 60.000 seconds so it's not because the switch purges the devices MAC address.
We haven't implemented QoS on the network (all Dell PowerConnect switches), yet.
Dell Business Support has of course been contacted (service level is 4 hours Gold), but they would not help us because they currently don't support VoIP on Dell switches!
Any advice is much appreciated!
Thanks in advance.
/MiGe
No Events found!
tester25
626 Posts
0
February 5th, 2007 20:00
MiGe
16 Posts
0
February 8th, 2007 07:00
* Portfast / Fast Link is enabled on all ports except up/down-link ports (spanning-tree portfast)
* All normal ports (where there's a PC or server connected to) are now configured as "General" (switchport mode general) and our Data VLAN is "attached" to all untagged packets (switchport general pvid 8)
* All up/down-link ports are now configured as "Trunk" (switchport mode trunk) with our Data VLAN as native (switchport trunk native vlan 8) and our VoIP VLAN as allowed (switchport trunk allowed vlan add 10)
Message Edited by MiGe on 02-08-2007 10:10 AM