By: Cory Carter user 21 Mar 2018 at 10:41 a.m. CDT

1 Response
Cory Carter gravatar
Hello, I'm looking into create a 3 - instance Redis-Sentinel cluster (1 master, 2 slaves) for HA/Failover purposes and am currently adding the appropriate JSON to the appliance fields according to [this doc](https://gluu.org/docs/ce/3.1.2/installation-guide/cluster/). My question is that, for "oxIDPAuthentication" and "oxCacheConfig", would including the slave nodes in the configuration work properly (i.e., with slave nodes being read-only, would hitting a node during a write operation autonomously pass this operation to a master node, or would it fail whenever this occurred?). Is there any suggestion for proper setup? Thanks, Cory

By Michael Schwartz Account Admin 21 Mar 2018 at 11:14 a.m. CDT

Michael Schwartz gravatar
Do you really need to cluster? If your SLA is 99.9%, that's 8 hours of downtime per year. If you do a good job monitoring, and detect an outage quickly, you can restore a new instance of the Gluu Server in that time. We recommend clusters when your SLA is 99.99% or higher. Most universities don't need this high of an SLA. And if you really do need it, shouldn't you have a support contact? And if you have a support contract, you can use cluster manager, which will make it easier for you to deploy clusters, and easier for us to support you. You can use Cluster manager for development or QA without a license. I would suggest that path. For more information on CLuster Manager, see the [documentation](https://gluu.org/docs/#cluster-manager).