topic RE: DHCP Client IP Address Loss on Stack Failover in ExtremeSwitching (EXOS/Switch Engine)

DHCP Client IP Address Loss on Stack Failover

Karl_Witthuhn — Tue, 20 Oct 2015 22:15:00 GMT

I have a configuration where I have a set of two stacked X440-48t switches (4 physical switches, 2 sets of 2-stacks). SW1 and SW2

I have a DHCP server connected to SW1 port 1:1

SW1 port 1:48 <-> SW2 port 1:48
SW1 port 2:48 <-> SW2 port 2:48

Here's the relevant configuration:

Firmware: 15.6.1.4 on both switches
stpd disabled on both switches

SW1:
enable mac-lockdown-timeout ports 1:48
configure mac-lockdown-timeout ports 1:48 aging-time 60
enable sharing 1:48 grouping 1:48, 2:48 algorithm address-based L2
enable dhcp vlan Default

SW2:
enable mac-lockdown-timeout ports 1:48
configure mac-lockdown-timeout ports 1:48 aging-time 60
enable sharing 1:48 grouping 1:48, 2:48 algorithm address-based L2
enable dhcp vlan Default

So the problem happens when I kill the power to the stacking master on SW2 - when I kill the power, the DHCP server can no longer reach SW2. I have a console open to both switches for SW2, I can issue a 'disable dhcp vlan default' and 'enable dhcp vlan default' on the new master switch (the one that remained powered on) and I can then reach the switch again.

I'm wondering if there's a way to automate this without having to console into the failed over switch and forcing a new DHCP request from the master switch.

Any help is appreciated!

Thanks.

RE: DHCP Client IP Address Loss on Stack Failover

Patrick_Voss — Tue, 20 Oct 2015 22:23:00 GMT

Does the same behavior happen when you simply unplug or disable port 1:48 on SW2? are you continuously pinging the DHCP address and it goes completely down or are you simply trying to connect to the address from the other switch, using telnet?

RE: DHCP Client IP Address Loss on Stack Failover

Karl_Witthuhn — Tue, 20 Oct 2015 22:36:00 GMT

Yeah - both SW1 and SW2 are a pair of stacked X440 switches (2 X440's in each stack)

The DHCP server is connected to SW1 port 1:1

There's a static port-channel between SW1 <-> SW2 (ports 1:48 & 2:48 on both switches)

Both switches have a DHCP entry on the DHCP server using the stack mac-address (02:04:XX:XX:XX:XX instead of the individual switch mac-address of 00:04:XX:XX:XX:XX) because when the switches are stacked and DHCP is enabled, it will use the stacked mac-address instead of individual mac-address.

I can have a continuous ping from the DHCP server -> SW2 and as soon as I kill the master switch on SW2 (ports 1:X), the ping is lost to SW2 and will never return unless I console into SW2 (ports 2:X) and disable -> enable DHCP on vlan default.

I'm trying to connect to SW2 via telnet from the DHCP server after a partial stack failure, so I need the IP address to remain reachable without console access if at all possible.

I'm wondering if there's a special script or method to automatically run a series of commands upon a stack failure. That way I could script it out so that if a switch became a master switch in a stack, it could run 'disable dhcp vlan default' and then 'enable dhcp vlan default' since that seems to fix my problem via console.

RE: DHCP Client IP Address Loss on Stack Failover

Patrick_Voss — Tue, 20 Oct 2015 23:52:00 GMT

Hi Karl,

I don't believe this is normal behavior. I would reach out to GTAC and open a case. If the proper entitlement is added to this device we may be able to attempt a replication and look for a proper fix.