Add an alert when it looks likely we will exhaust the shard capacity of our existing data nodes in the next 2 weeks. If it's hard to predict when this will be at a first approximation let's assume this alert should fire at 90% shard usage.
We want this alert so that we don't fully exhaust the shard count and then are unable to make indices for new wikis.
Exceeding this number is very inconvenient since we then have to go back and recreate them.
On this alert being triggered we need to provision a new Elasticsearch data node; we should also check at this time that we aren't seeing the master node using more than 75% of their heap. If they are we need to add more Memory to the master nodes.
In order to take either of these actions we may need to add more kubernetes nodes to the cluster.