markwalkom/ES restart + recovery

## ES restart + recovery
## For a node

This tells ES to not move shards around if a node/nodes drops out of the cluster;
`curl -XPUT eshost.example.com:9200/_cluster/settings -d '{"transient":{"cluster.routing.allocation.enable": none}}'`
`sudo service elasticsearch restart`
`curl -XPUT eshost.example.com:9200/_cluster/settings -d '{"transient":{"cluster.routing.allocation.enable": all}}'`
When you are done and the node/nodes are back, they will reinitialise their local shards and you're cluster should be back to green as fast as the local node can work.

## For a cluster

You can also speed a full cluster restart by settings the following config settings - *gateway.recover_after_nodes* and *gateway.expected_nodes* - preferably to the number of nodes in the cluster. This way when a cluster comes back it won't start recovery until it can see the numbers specified and when it does it will use local data as much as possible.

Finally you can set *index.recovery.concurrent_streams*, this requires a bit of testing to make sure you don't overwhelm your node, but if you can increase that from the default of 3 to something higher.
	## For a node

	This tells ES to not move shards around if a node/nodes drops out of the cluster;
	`curl -XPUT eshost.example.com:9200/_cluster/settings -d '{"transient":{"cluster.routing.allocation.enable": none}}'`
	`sudo service elasticsearch restart`
	`curl -XPUT eshost.example.com:9200/_cluster/settings -d '{"transient":{"cluster.routing.allocation.enable": all}}'`
	When you are done and the node/nodes are back, they will reinitialise their local shards and you're cluster should be back to green as fast as the local node can work.

	## For a cluster

	You can also speed a full cluster restart by settings the following config settings - gateway.recover_after_nodes and gateway.expected_nodes - preferably to the number of nodes in the cluster. This way when a cluster comes back it won't start recovery until it can see the numbers specified and when it does it will use local data as much as possible.

	Finally you can set index.recovery.concurrent_streams, this requires a bit of testing to make sure you don't overwhelm your node, but if you can increase that from the default of 3 to something higher.