Don't Go Into ElasticSearch Production Without This Checklist

Elasticsearch is a great tool to provide fast and powerful search services to your web sites or applications, but care should be taken when moving from development to production. By following the checklist below, you can avoid some issues that may arise if you use development settings in a production environment!

Configure Your Log and Data Paths

To minimize the chances of data loss in a production environment, it is highly recommended that you change your log and data paths from the default paths to something that is less likely to be accidentally overwritten.

You can make these changes in the configuration file (which uses YAML syntax) under path, as in the following example, which uses suggested production paths from the Elasticsearch team:

Suggested settings for the log and data paths. Source: Elasticsearch.

Configure Your Node and Cluster Names

When you are looking for a node or a cluster, it is a good idea to have a name which describes what you will need to find and separates one from another.

The default cluster name of "elasticsearch " could allow any nodes to join the cluster, even if this was not intended. Thus, it is a good idea to give the cluster a distinct identifier instead.

The default node names are chosen randomly from a set of roughly 3000 Marvel character names. While this wouldn't be so bad for a node or two, this could get quite confusing as you add more than a few nodes. The better option is to use a descriptive name from the beginning to avoid potential confusion as nodes are added later.

Configure Memory Settings

Memory swapping used on systems could potentially cause the elasticsearch process to be swapped, which would not be good while running in production. Suggestions from the Elasticsearch team to fix this include disabling swapping, configuring swapping to only run in emergency conditions, or (for Linux/Unix users) using mlockall to try to lock the address space of the process into RAM to keep it from being swapped.

Configure Virtual Memory Settings

Elasticsearch indices use mmapfs/niofs, but the default mmap count on operating systems can potentially be too low. If so, you will end up with errors such as "out of memory" exceptions. To fix this, you can up the limit to accommodate Elasticsearch indices. The following example shows how the Elasticsearch team recommends increasing this limit on Linux systems (run the command as root):

Suggested command to increase the mmap count for Linux systems. Source: Elasticsearch.

Ensure Elasticsearch is Monitored

It is a good idea to monitor your Elasticsearch installation so that you can see the status or be alerted if or when something goes wrong. A service such as Happy Apps can provide this type of monitoring for you (and can monitor the rest of your app as well).

Get ElasticSearch in the Cloud

When you launch your application that uses Elasticsearch, you will want reliable and stable database hosting. Morpheus Virtual Appliance is a tool that allows you manage heterogeneous databases in a single dashboard.