Configuring and tuning Elasticsearch¶
We strongly recommend to use a dedicated Elasticsearch cluster for your Graylog setup. If you are using a shared Elasticsearch setup, a problem with indices unrelated to Graylog might turn the cluster status to yellow or red and impact the availability and performance of your Graylog setup.
Configuration of graylog-server nodes¶
The most important settings to make a successful connection are the Elasticsearch cluster name and the discovery mode. Graylog is able to discover the Elasticsearch nodes using multicast. This is great for development and proof of concepts but we recommend to use classic unicast discovery in production.
You need to tell
graylog-server which Elasticsearch cluster to join. The Elasticsearch cluster default name is elasticsearch
and configured for every Elasticsearch node in its
elasticsearch.yml configuration file with the setting
Configure the same name in every
We recommend to call the cluster
graylog-production and not
elasticsearch.yml file is typically located in
The default discovery mode is multicast. Graylog will try to find other Elasticsearch nodes automatically. This usually works fine when everything is running on the same system but gets problematic quickly when running in a bigger network topology. We recommend to use unicast for production setups. Configure Zen unicast discovery in Graylog with the following lines in your configuration file:
# Disable multicast elasticsearch_discovery_zen_ping_multicast_enabled = false # List of Elasticsearch nodes to connect to elasticsearch_discovery_zen_ping_unicast_hosts = es-node-1.example.org:9300,es-node-2.example.org:9300
Also make sure to configure Zen unicast discovery in
the Elasticsearch configuration file by adding the
discovery.zen.ping.unicast.hosts setting with the
list of Elasticsearch nodes to
discovery.zen.ping.multicast.enabled: false discovery.zen.ping.unicast.hosts: ["es-node-1.example.org:9300" , "es-node-2.example.org:9300"]
The Elasticsearch default communication port is 9300/tcp (not to be confused with the HTTP interface running on port 9200/tcp by default).
The communication port can be changed in the Elasticsearch configuration file (
elasticsearch.yml) with the configuration setting
Make sure that Elasticsearch binds to a network interface that Graylog can connect to (see
Configuration of Elasticsearch nodes¶
Disable dynamic scripting¶
Make sure to add
script.disable_dynamic: true to the
elasticsearch.yml file to disable the dynamic scripting feature and
prevent possible remote code executions.
Control access to Elasticsearch ports¶
Since Elasticsearch has no authentication mechanism at time of this writing, make sure to restrict access to the Elasticsearch ports (default: 9200/tcp and 9300/tcp). Otherwise the data is readable by anyone who has access to the machine over network.
Open file limits¶
Because Elasticsearch has to keep a lot of files open simultaneously it requires a higher open file limit that the usual operating system defaults allow. Set it to at least 64000 open file descriptors.
Graylog will show a notification in the web interface when there is a node in the Elasticsearch cluster which has a too low open file limit.
Read about how to raise the open file limit in the corresponding Elasticsearch documentation page.
It is strongly recommended to raise the standard size of heap memory allocated to Elasticsearch. Just set the
variable to for example
24g to allocate 24GB. We recommend to use around 50% of the available system memory for Elasticsearch (when
running on a dedicated host) to leave enough space for the system caches that Elasticsearch uses a lot.
Graylog is already setting specific configuration per index it creates. This is enough tuning for a lot of use cases and setups. A more detailed guide on deeper tuning of Elasticsearch is following.
Cluster Status explained¶
Elasticsearch provides a classification for the cluster health:
The red status indicates that some or all of the primary shards are not available. In this state, no searches can be performed until all primary shards are restored.
The yellow status means that all of the primary shards are available but some or all shard replicas are not.
With only one Elasticsearch node, the cluster state cannot become green because shard replicas cannot be assigned. This can be solved by adding another Elasticsearch node to the cluster.
If the cluster is supposed to have only one node it is okay to be in the yellow state.
The cluster is fully operational. All primary and replica shards are available.