Last week I've installed a new stacking unit in our network. Before last week, we had only one core switch, which is obviously a very large SPOF. This was the first time I've configured a stack in a network. It all works great, but I think that our failover time could be faster, in case of one switch is going down.
It takes about 30 seconds before a client is connected again to all the network resources. 30 seconds typically sounds like spanning tree may be causing this, but I am not sure of that. I tested the failover time by unplugging one switch of power, which simulates a UPS of other power failure (of course, each switch is connected to another UPS).
Our configuration / network looks like this: Two 6224's stacked Most of our access switches are 35xxP series
The access switches are connected directly into the core switch over two links divided to each switch of the stack. Those two links are configures in port channel mode to increase bandwith and failover automaticity.
For example access switch one G3 is connected to 1/g6 of our stack and G4 is connected to 2/g6.
The partial configuration:6224 stack:
stack
member 1 1
member 2 1
exit
switch 1 priority 12
switch 2 priority 10
~~~~~
interface ethernet 1/g6
channel-group 6 mode auto
exit
interface ethernet 2/g11
channel-group 6 mode auto
exit
~~~~~
interface port-channel 6
service-policy in
spanning-tree guard root
switchport mode general
switchport general allowed vlan add 12,20 tagged
switchport general allowed vlan add 1 tagged
exit
Spanning tree:
Spanning tree :Enabled - BPDU Flooding :Disabled - Portfast BPDU filtering :Disabled - mode :rstpCST Regional Root: 10:00:D0:67:E5:75:6D:D9
Regional Root Path Cost: 0
ROOT ID
Address D0:67:E5:75:6D:D9
This Switch is the Root.
Hello Time 2 Sec Max Age 20 sec Forward Delay 15 sec TxHoldCount 6 sec
Access switches:
spanning-tree mode rstp
interface range ethernet e(1-24)
spanning-tree portfast
exit
interface range ethernet e(1-24)
spanning-tree bpduguard
exit
interface port-channel 1
switchport mode general
switchport general allowed vlan add 12,20
exit
~~~~
interface range ethernet g(3-4)
channel-group 1 mode auto
exit
One idea I had was to disable spanning tree, but spanning tree should never be disabled because there is always a possibility to create a loop by accident. For testing purposes I've disabled STP on the port-channel on both sides, but that didn't speed up the failover time.
Does anyone see optimizations to speed up the failover time? In my test environment I had a failover time of almost zero our one ping transmit failure (though I had the same config, but it seems not to be...). It would be great I can achieve this time again.