Dynamically scoping Redis with AWS Elasticache to use the local availability zone. Batch writing is used for the shipper to reduce usage costs.

logstash-elasticache.pp

$redis_host = $::ec2_placement_availability_zone ? {
  'us-west-1a' => 'logging-redis-a.<redacted>.0001.usw1.cache.amazonaws.com',
  'us-west-1b' => 'logging-redis-b.<redacted>.0001.usw1.cache.amazonaws.com'
}

logstash::output::redis { 'redis-shipper':
  batch     => true,
  host      => $redis_host,
  key       => 'logstash',
  data_type => 'list'
}

logstash::input::redis { 'redis-indexer':
  host      => $redis_host,
  key       => 'logstash',
  data_type => 'list'
}

Valuable insight from this config by @hatt:

you could do something like set a variable (maybe $region_id) to just be a or b or something
and interpolate it for entire hostgroups as an array so if you had 6 redis servers, 3 in each 
AZ for whatever reason, just pass the one var and popular the host array

btw, I can confirm that using only the local AZ in LS/redis/elasticache saves $1/100GB of 
traffic

what'd be really cool is if you say, had a listener for AZ health or something, and if a 
particular AZ had elasticache issues you could swap LS out temporarily to a different 
node by triggering a puppet run, it'd cost you a bit but it;d prevent LS backing up :)

Thanks Mr Hatt!