.
I will start here with STP.
As covered in one of my earlier posts, When configuring Spanning Tree on Force10 switches, you should accommodate whether your fabric is isolated, or inter-connects to other switches.
The implications are as follows:
On an isolated fabric or isolated switch segments, like the ones often used in iSCSI deployments, the preferred STP mode in my opinion should always be RSTP .1W. RSTP’s timers, states and topology change handling is far superior and efficient in comparison to STP .1D. Ofcourse, one could argue that by virtue of its nature, an isolated fabric (not segment) does not connect to another switch, so what is the point of having spanning tree in operation.
the answer is that it just comes down to good practice. I have not yet come across a situation where issues were caused due to a bridging application being run on an edge port server, but i have come across situations where someone inadvertently plugged in a switch, which was a part of network’s active forwarding topology, into an isolated fabric via multiple links. Unfortunately, STP was explicitly disabled on the fabric. it was some time before issues started to manifest, but when the did, the whole network was brought down to its knees in no time.
on a fabric that interconnects to other switches, the first thing to obviously account for, is the model/brand of the other switches. what you want to ensure is a network without broken STP topologies/domains. If the other switches are Force10 as well, great! if not, some thought has to be invested upfront into STP interoperability. if they happen to be Cisco, (and you are not keen on using MSTP), the alternate available to you is to use Force10’s PVST mode, which is compatible with Cisco’s Per VLAN implementation of STP. This would enable you to have spanning tree forwarding topologies on a per VLAN basis that span the entire network.
PVST is the IEEE nomenclature, and PVST+ is Cisco’s equivalent to the IEEE’s PVST version. Cisco switch runs a separate STP instance for each VLAN (thus called PVST), while standard IEEE compliant switches run just one instance of STP shared by all VLANs. Force10’s PVST is based on STP .1d, and uses the .1s (MSTP) costs as the default values/timers. There are a couple of other alternates to make STP inter-operate but i have covered them in a separate post, which you can find here: https://hasanmansur.com/2012/10/15/cisco-and-force10-stp-spanning-tree-interoperability/
you can further tune STP, but you have to be very careful with this step or you could be introducing instabilities in your network.
First, to configure STP in any mode, you will use:
protocol spanning-tree rstp
or
protocol spanning-tree pvst
no disable
Yes, you have to issue a “no disable” under the protocol stanza, to start the STP process. you should further configure bridge priorities for deterministic root placement.
bridge-priority xxxx
With extreme care & caution, you can tune RSTP timers to their minimum to get the fastest possible convergence, (but only do this for RSTP .1w), & never in very large networks.
forward-delay 4 (Default: 15)
max-age 6 (Default: 20)
hello-time 1 (Default: 2)
The hello time is encoded in BPDUs in increments of 1/256ths of a second. The standard minimum hello time in seconds is 1 second, which is encoded as 256. When millisecond hellos are configured, the default hello interval of 2 seconds is still used for edge ports; the millisecond hello interval is not used. Hello time 1 offers interoperability at the cost of negligibly slower convergence (instead of “hello-time milli-second 50”).RSTP Fast hellos (hello-time milli-second interval Range: 50 – 950 milliseconds) can be used for sub second link detection.
As for the VLAN configuration, the commands are as follows:
Config)# interface vlan z / interface range VLAN [1 – 4094]
Config)# no shutdown
Port membership is configured under the VLAN interface. On Cisco, you have Trunk and Access modes, but in Force10, Ports are defined as tagged or untagged members of the respective VLAN. This way of configuring port membership makes more sense actually. for e.g. if you have a port on Cisco device defined as trunk, carrying VLANs 10, 20 and 30, with native vlan 100, you can accomplish the same by defining the VLAN membership on force10 to be tagged for VLANs 10, 20 and 30, and untagged for VLAN 100, for that certain port.
To add an interface to a VLAN, it must be in Layer 2 mode. Ensure you have accomplished this by issuing the
switchport
command under the port stanza. with this done, the membership can be defined as follows:
(conf-if-vlan)# tagged interface
(conf-if-vlan)# untagged interface
Lastly, to introduce portfast equivalent on Force10 switches, use the following:
spanning-tree [rstp/pvst] edgeport [bpduguard | shutdown-on-violation]
edgeport feature in RSTP, is the equivalent of portfast feature in STP. It skips Listening/Learning stage, gains 30 sec on forwarding.
.
Hasan Mansur 