Cloaking risk for load balanced Drupal site where new content may take a while to get to all servers

Question

I want to use Load Balancer(2 nodes) for one my Drupal site. That means editor can't edit the site since edits will go to any node. Secondly comments can also be saved in either node.

Even if we edit the site locally and migrate the changes to all the nodes, sometimes there can be issue in which can one of the nodes may not be fully updated. However there is no option for comments in this case

When Google crawls the content it can find different content at different times from same url.

How to handle cloaking issue arising due to use of Load Balancer?

Answer 1

What you are describing is not cloaking. Cloaking is serving different content to Googlebot based on the user-agent header. In the senario you describe, both users and Googlebot have the chance of seeing incorrect content. You are not trying to be sneaky and fool Googlebot by showing it something different from users. There is no risk of Google penalizing your site because of this.

The behavior you describe is not ideal for users. You should be able to run Drupal behind a load balancer without such issues. To do so:

• Centralize your database
• Use a shared NFS mount for the files that Drupal writes
• Use a cache backend that uses Memcached or Redis (rather than local disk cache)
• Use a script to upgrade all your Drupal instances at once
• Ensure that each Drupal cron job is only running once (installed on a single server)

---

                          ┌─→ Web server 1 ↘
Browser ─→ Load balancer ─┼─→ Web server 2 → Database
                          └─→ Web server 3 ↗

References:

• Using a load balancer or reverse proxy | Drupal.org
• Server Scaling | Managing site performance and scalability | Drupal guide on Drupal.org
• Load balancing multiple horizontal drupal instances - Drupal Answers
• What is the proper way to set up Drupal 7 on load balanced servers so that CSS and JS files are in sync? - Drupal Answers

Answer 2

I believe you have two databases corresponding to the two nodes. Maybe you can move the database to a separate machine or a cloud service (like AWS RDS for example). After such a setup, any change should reflect in both the nodes even if you update either of the nodes.

If you do not want to set up a separate database server, maybe you can try some kind of real-time syncing between the database instances. I believe solutions will exist for such bidirectional syncing but can be complex and resource-demanding.