A/B testing entire website VS Seo issues

[SeoMartin1]

I'm familar with A/B testing variations of a page but I'd like to A/B test a new designs version of a e-commerce site. I´m wondering about the best way to test with SEO concerns... this is what I´ve in mind right now, any suggestion?

1. Use parameters to make version B different from A version.
2. Redirect 50% of the users with 302 ( or javascript would be a better way?)
3. Use noindex in the B pages.
4. Use rel=canonical in the B pages pointing to A version.
5. In the end use 301 redirect to all B pages to A urls.

PS: We can´t use subdomain and i don´t wanna use robots.txt file to protect the new design from competitors.

I´d love any suggestions and tips about it - thanks folks

[ecommercebc]

My main experience with A/B is via Google Experiments and it looks like they highlight option 4:

To ensure that serving your variation pages does not have a negative impact on your site's SEO rankings, you can use therel="canonical" link attribute on your variation pages. rel="canonical" is a signal to search engines that the content of your variation pages is essentially the same as that of your original page, and that you would prefer that search engines index your original page rather than the variations you're using in your experiment.

Source: https://support.google.com/analytics/answer/2613318

If you roll out rel=canonical correctly then this seems to be your safest bet in Google's eyes. My main experience is with split-testing single pages, but theoretically this should work out okay with the correct use of canonical. If you're really not sure you could roll out a mini-test with a couple of key pages first?