Merge pull request #2472 from carmstrong/clarify_resouce_requirements

docs(installing_docs): add cluster size to system requirements

Author: carmstrong

docs/installing_deis/system-requirements.rst

@@ -12,17 +12,38 @@ will move around the cluster onto healthy hosts as hosts leave the cluster for v
 (failures, reboots, autoscalers, etc.). Because of this, you should have ample spare resources on
 any machine in your cluster to withstand the additional load of running services for failed machines.
 
-Machines must have:
+Resources
+---------
 
-* At least 4GB of RAM (Deis uses 2 - 2.5GB, plus room for failover and deployed applications)
+Deis components consume approximately 2 - 2.5GB of memory across the cluster, and approximately
+30GB of hard disk space. Because each machine should be able to absorb additional load should a
+machine fail, each machine must have:
+
+* At least 4GB of RAM (more is better)
 * At least 40GB of hard disk space
 
+Note that these estimates are for Deis and CoreOS only, and there should be ample room for deployed
+applications.
+
 Running smaller machines will likely result in increased system load and has been known to result
 in component failures, issues with etcd/fleet, and other problems.
 
-If running multiple (at least 3) machines of an adequate size is unreasonable, it is recommended to
+Cluster size
+------------
+
+For scheduling to work properly, clusters must consist of at least three nodes and always
+have an odd number of members. This is mostly because the underlying CoreOS cluster must always
+be able to obtain a quorum (see `optimal etcd cluster size`_). Additionally, the :ref:`Store`
+component keeps three replicas of stored data, so it requires at least three nodes.
+
+.. important::
+
+    Deis clusters of less than three nodes are unsupported.
+
+If running multiple (at least three) machines of an adequate size is unreasonable, it is recommended to
 investigate the `Dokku`_ project instead. Dokku is `sponsored`_ by Deis and is ideal for environments
 where a highly-available distributed system is not necessary (i.e. local development, testing, etc.).
 
 .. _`dokku`: https://github.com/progrium/dokku
+.. _`optimal etcd cluster size`: https://github.com/coreos/etcd/blob/master/Documentation/optimal-cluster-size.md
 .. _`sponsored`: http://deis.io/deis-sponsors-dokku/