Sugar 7.7 on AWS behind load balancer

Question

I'm trying to set up a Sugar 7.7 install on AWS and make use of autoscaling so that when the sales system this is integrated with runs its scheduled job to update customer and other account information, there's CPU cycles available to continue doing other work. I've made a couple of attempts, but it seems that while I can log in, the token becomes invalid as soon as the dashboard pages tried to load. The reading I've managed to do so far seems to indicate this is due to some hashing algorithm that requires the Sugar files on every web host to be identical, and that the differences are due to the cache. The only solution I've seen mentioned is to use NFS, but this won't work easily on AWS, and can get very complex when you try to set it up to be fault tolerant.

Is there any how-to out there that can help, or does anyone here have any advice or information they can share on how I can successfully set this up?

Thanks

Jason Gilpin · Accepted Answer

Hi Jason,

Based on your description, it sounds like some options would be:

Use Redis or Memcached for PHP Session storage -- We use Redis clusters in our Reference Builds
Enable sticky sessions at the load balancer to bind sessions to a specific instance for the duration of the session -- Could impact the effectiveness of the load balancer since existing sessions do not get "re-balanced"

Thanks,

Dan