Should I enable Spanning Tree ?

[jdufour]

Hello,

I may be a little old school. In my experience and training, Spanning Tree Protocol (STP) has always been considered mandatory. We are reconfiguring our network and the topic of STP is on the table. Per the HP documentation, STP is only recommened when you know you have a physical loop. See link:

ftp://ftp.hp.com/pub/networking/software/59903016e7_ch13.pdf

page 13-3.

I find this recommendation to be odd as I would never intentionaly build a newtork which has physical loops. We currently are not running STP and have no issues. Should we enable STP during our network reconfiguration? Pros? Cons?  Thanks for your time! -  John

[efa-pFunk]

Hi, 

I think they mean with loops, such redundant links.
In some situation, this is usefull, just to create a higher availability.

In fact, if you have a flat network with one core-switch and just a bunch of access/edge switches, which are only connected with one link, STP might be overrated.

If you plan to have more than one core-switch - e.g. you have 2 datacenters or more - and you plan to connect every switch directly to the core-switches, you have 2 different ways for each switch, creating physical loops.
Unless there are configurations like VSF or LACP or whatever, it might be usefull to check, if STP is needed.

With our last vendor we had problems with STP, so we had to manually build rules to cut down links on some switches to prevent them of building loops and than turned STP off - didn't bring in the comfort as we had with STP before, because it was not fully automated to bring back each access-switch while testing the scenarios, but didn't help...

[parnassus]

I recall a very nice statement made by another HPE Community user (I should check who actually is...I don't recall now), it is more or less: "Always plan a network topology to function properly even without (R)STP but leave (R)STP feature enabled".

LACP Trunks between Switches work (R)STP without issues.

Generally if, as written above, Core adopts VSF/IRF implementations (so, in both cases, a form of logical front/back-plane stacking) you should not worry about LACP/Non-Protocol Port Trunking and (R)STP feature because port trunks can be terminated on multiple stacked switches (in so you add path resiliency and traffic balancing) without particular issues (there is no loop because the logical switch is one and one only even if you are working with different physical units).

So it's a matter of implementing/planning a good network topology keeping an eye open especially on edge layer (this is the layer where user generated connection loops can happen more often with respect to what can happen at the core/distribution layers). 

[Vince-Whirlwind]

I've always thought HP's message was that STP - disabled by default - is an optional extra.

To me, it's mandatory configuration.

Also, to echo the comment above, I generally would not expect STP to be blocking any links unless something had gone awry.