Review the installation requirements below, and then check out available options to install and deploy InfluxDB Enterprise. For an overview of the architecture and concepts in an InfluxDB Enterprise cluster, review Clustering in InfluxDB Enterprise.
Requirements for InfluxDB Enterprise clusters
InfluxDB Enterprise clusters require a license. To use a license key, all nodes in the cluster must be able to contact https://portal.influxdata.com via port
80 or port
443. If nodes in the cluster cannot communicate with https://portal.influxdata.com, you must use the
license-path configuration setting. For more information, see Enterprise license settings.
Nodes attempt to download a new license file for the given key every four hours. If a node cannot connect to the server and retrieve a new license file, the node uses the existing license file. After a license expires, nodes have the following grace periods:
- If InfluxDB daemon (
influxd) starts and fails to validate the license, the node has a 4-hour grace period.
influxdstarts and validates the license, and then a later license check fails, the node has a 14-day grace period.
Frequently overlooked requirements
The following are the most frequently overlooked requirements when installing a cluster.
- Ensure connectivity between machines
- Synchronize time between hosts
- Use SSDs
- Do not use NFS or NFS-based services
- Disable swap
- Use three and only three meta nodes
- Meta and data nodes are fully independent
- Install Chronograf last
Ensure connectivity between machines
All nodes in the cluster must be able to resolve each other by hostname or IP, whichever is used in the configuration files.
For simplicity, ensure that all nodes can reach all other nodes on ports
If you alter the default ports in the configuration file(s), ensure the configured ports are open between the nodes.
Synchronize time between hosts
InfluxDB Enterprise uses hosts’ local time in UTC to assign timestamps to data and for coordination purposes. Use the Network Time Protocol (NTP) to synchronize time between hosts.
Clusters require sustained availability of 1000-2000 IOPS from the attached storage. SANs must guarantee at least 1000 IOPS is always available to InfluxDB Enterprise nodes or they may not be sufficient. SSDs are strongly recommended, and we have had no reports of IOPS contention from any customers running on SSDs.
Do not use NFS or NFS-based services
For disk storage, use block devices only. InfluxDB Enterprise does not support NFS (Network File System)-mounted devices or services such as AWS EFS, Google Filestore, or Azure files.
To avoid potential disk contention when InfluxDB is under high load, disable swap in your operating system settings.
Use three and only three meta nodes
Although technically the cluster can function with any number of meta nodes, the best practice is to ALWAYS have an odd number of meta nodes. This allows the meta nodes to reach consensus. An even number of meta nodes cannot achieve consensus because there can be no “deciding vote” cast between the nodes if they disagree.
Therefore, the minimum number of meta nodes for a high availability (HA) installation is three. The typical HA installation for InfluxDB Enterprise deploys three meta nodes.
Aside from three being a magic number, a three meta node cluster can tolerate the permanent loss of a single meta node with no degradation in any function or performance. A replacement meta node can be added to restore the cluster to full redundancy. A three meta node cluster that loses two meta nodes will still be able to handle basic writes and queries, but no new shards, databases, users, etc. can be created.
Running a cluster with five meta nodes does allow for the permanent loss of two meta nodes without impact on the cluster, but it doubles the Raft communication overhead.
Meta and data nodes are fully independent
Meta nodes run the Raft consensus protocol together, and manage the metastore of all shared cluster information: cluster nodes, databases, retention policies, shard groups, users, continuous queries, and subscriptions.
Data nodes store the shard groups and respond to queries. They request metastore information from the meta group as needed.
There is no requirement at all for there to be a meta process on a data node, or for there to be a meta process per data node. Three meta nodes is enough for an arbitrary number of data nodes, and for best redundancy, all nodes should run on independent servers.
Install Chronograf last
Chronograf should not be installed or configured until the InfluxDB Enterprise cluster is fully functional.
Set up monitoring
Monitoring gives you visibility into the status and performance of your cluster. See “Monitor the InfluxData Platform” for information on setting up monitoring for your InfluxDB Enterprise installation.
Was this page helpful?
Thank you for your feedback!
Support and feedback
Thank you for being part of our community! We welcome and encourage your feedback and bug reports for InfluxDB and this documentation. To find support, use the following resources:
InfluxDB Cloud and InfluxDB Enterprise customers can contact InfluxData Support.