Elasticsearch CollectD plugin

A CollectD plugin to collect Elasticsearch stats and metrics. Uses CollectD's Python plugin.

Installation

1. Place the elasticsearch_collectd.py file into a directory on the host. The recommended directory is /usr/share/collectd/collectd-elasticsearch.
2. Configure the plugin (see below).
3. Restart collectd.

Requirements

• collectd 4.9+
• Elasticsearch 1.x or newer.
• python v2.6 or newer (https support requires v2.7.9)

Configuration

• See elasticsearch.conf
• The plugin will automatically determine the version of Elasticsearch you are running as well as the cluster name.
• Per-index and cluster stats can be disabled if needed; they are enabled by default. These settings are collected only on master eligible nodes.
• If you are running the Elasticsearch plugin via a collectd deployment within a container, please configure the Host and Port values inside of the 20-elasticsearch.conf file that correspond to the desired Elasticsearch instance.

ex:

   <Module "elasticsearch_collectd">
       Host "XXX.XXX.XXX.XXX"
       Port "XXXX"
   </Module>

Metrics

Node stats

• Documents (total docs & deleted docs)
• Store size
• Indexing (total, time, total delete, delete time)
• Get (total, time, exists total, exists time, missing total, missing time)
• Search (total query, total time, total fetch, total fetch time)
• JVM uptime
• JVM memory (heap committed, heap Used, non heap committed, non heap used)
• JVM threads (count & peak)
• JVM GC (time & count)
• Transport stats (server open, RX count, RX size, TX count, TX size)
• HTTP stats (current open & total open)
• OS stats (CPU percent, file descriptors)
• Thread pool stats (generic, index, get, snapshot, merge, optimize, bulk, warmer, flush, search, refresh)
• Cache (field eviction, field size, filter evictions, filter size)
• JVM collectors
• FLush (total count, total time)
• Merges (current count, current docs, current size, merge total size, docs a time)
• Refresh (Total & Time)

Index stats

• Transaction log (size, number of operations)
• Most of the common stats per index and per primary vs. total.

Cluster stats

• Shard stats (active, initializing, relocating, unassigned, primaries)
• Nodes (total, data nodes)