I'm still struggling with this, if someone could help it would be appreciated. Virtual IP's just dont seem to be transferable through these switches and i'm sure it's to do with the ARP's. We have a number of older 3COM switches that work fine but these Dells just dont want to play the game.
I just ran accross this post, having the same problem... I have a virtualization cluster (XenServer), with vlan trunks into the 6248. When I migrate a VM, pings to the VM stop for upwards of 30 seconds, unless the VM generates some traffic of its own (IE, hop on the console and ping out).
Using tcpdump on the firewall/router, I can see XenServer sending out the gratuitous arp -- I just think the PowerConnect is ignoring those arp replies.
I do not see any feature/options for gratuitous arp on the 62xx model switch.
If you have spanning tree enabled on the switch it may be influencing the behavior on the port. You can configure portfast on the interface to help stabilize the spanning tree convergence time. When a device leaves a port and enters another it can kick off a spanning tree recalculation that would create a 30 second window of a nonresponsive port.
Actual config for the trunk-ports going to the two servers... As you can see, portfast is definately enabled. Does it matter that these are trunk ports? Do I need to do something to the VLAN in order to make it actuall process, rather than just forwarding, the gratuitous arp frames?
You may try using switchport mode general on the ports. It is a mix of access and trunk mode.
General Links consist of a combination of VLAN Trunk and Access Links.
General Links can have both tagged and untagged frames, However, all frames sent to a specific VLAN must be tagged. All untagged frames are sent to the native VLAN
.
The native VLAN still applies to the General LINK. While it is possible to have multiple untagged vlans on a General link, you can only have ONE (1) PVID. The PVID represents the native VLAN. While untagged traffic may be sent via several untagged VLANs, returning untagged traffic will only be received by the PVID and therefore will NOT be forwarded to a specific VLAN.
General links are mostly used today for legacy equipment. However, on the PowerConnect 62xx series switches, you must use General mode if you want to allow management traffic onto the switch over the PVID. If you use Trunk mode, you will not have the default VLAN on those ports. The ports will only allow tagged traffic.
My default VLAN is 1, which can't be trunked on these switches (not sure why?), so *nothing* is in that VLAN. All normal access ports are explicitly defined to be in vlan 2.
That said, I can try General mode and see what happens...
GordonCopestake
6 Posts
0
January 23rd, 2008 05:00
msturtz
4 Posts
0
October 5th, 2012 08:00
I just ran accross this post, having the same problem... I have a virtualization cluster (XenServer), with vlan trunks into the 6248. When I migrate a VM, pings to the VM stop for upwards of 30 seconds, unless the VM generates some traffic of its own (IE, hop on the console and ping out).
Using tcpdump on the firewall/router, I can see XenServer sending out the gratuitous arp -- I just think the PowerConnect is ignoring those arp replies.
Did anybody solve this?
-Matt-
DELL-Willy M
802 Posts
0
October 5th, 2012 09:00
I do not see any feature/options for gratuitous arp on the 62xx model switch.
If you have spanning tree enabled on the switch it may be influencing the behavior on the port. You can configure portfast on the interface to help stabilize the spanning tree convergence time. When a device leaves a port and enters another it can kick off a spanning tree recalculation that would create a 30 second window of a nonresponsive port.
The following example enables PortFast on 1/g5.
console(config)#interface ethernet 1/g5
console(config-if-1/g5)#spanning-tree portfast
Hope this helps,
Keep us updated if you can.
msturtz
4 Posts
0
October 5th, 2012 11:00
Actual config for the trunk-ports going to the two servers... As you can see, portfast is definately enabled. Does it matter that these are trunk ports? Do I need to do something to the VLAN in order to make it actuall process, rather than just forwarding, the gratuitous arp frames?
console#show running-config interface ethernet 1/g13
description 'host3-trunk'
spanning-tree portfast
switchport mode trunk
switchport trunk allowed vlan add 2,100-101,111,121,131,141,151,161
console#show running-config interface ethernet 1/g22
spanning-tree portfast
switchport access vlan 2
console#show running-config interface ethernet 2/g22
description 'host1-trunk'
spanning-tree portfast
switchport mode trunk
switchport trunk allowed vlan add 2,100-101,111,121,131,141,151,161
console#
DELL-Willy M
802 Posts
0
October 5th, 2012 12:00
You may try using switchport mode general on the ports. It is a mix of access and trunk mode.
General Links consist of a combination of VLAN Trunk and Access Links.
General Links can have both tagged and untagged frames, However, all frames sent to a specific VLAN must be tagged. All untagged frames are sent to the native VLAN
.
The native VLAN still applies to the General LINK. While it is possible to have multiple untagged vlans on a General link, you can only have ONE (1) PVID. The PVID represents the native VLAN. While untagged traffic may be sent via several untagged VLANs, returning untagged traffic will only be received by the PVID and therefore will NOT be forwarded to a specific VLAN.
General links are mostly used today for legacy equipment. However, on the PowerConnect 62xx series switches, you must use General mode if you want to allow management traffic onto the switch over the PVID. If you use Trunk mode, you will not have the default VLAN on those ports. The ports will only allow tagged traffic.
msturtz
4 Posts
0
October 5th, 2012 12:00
My default VLAN is 1, which can't be trunked on these switches (not sure why?), so *nothing* is in that VLAN. All normal access ports are explicitly defined to be in vlan 2.
That said, I can try General mode and see what happens...