Documentation

Back up and restore InfluxDB Enterprise clusters

This page documents an earlier version of InfluxDB Enterprise. InfluxDB Enterprise v1.9 is the latest stable version. View this page in the v1.9 documentation.

When deploying InfluxDB Enterprise in a production environment, we recommend that you prepare for unexpected data loss with a strategy for backing up and restoring your InfluxDB Enterprise clusters.

Plan your back up and restore strategy

For each of the following scenarios, plan a back up and restore strategy that suits your needs:

  • Disaster recovery due to unexpected events
  • Migrating data to new environments or servers
  • Restoring clusters to a consistent state
  • Debugging

To choose a strategy that best suits your use case, we recommend considering your volume of data, application requirements, acceptable recovery time, and budget. We offer the following methods for backup and restore:

Test your backup and restore strategy for all applicable scenarios.

Backup and restore utilities

Use backup and restore utilities to back up and restore InfluxDB data from any of the following:

  • Cluster
  • Database
  • Database and retention policy
  • Single shard
  • From InfluxDB Enterprise to InfluxDB OSS or OSS to Enterprise

Back up and restore data between influxd instances with the same versions or with only minor version differences. For example, you can backup from 1.7.3 and restore on 1.7.10.

Backup utility

Important: Save backups to a new cluster or existing cluster with a different database name.

Backup creates a copy of the metastore and shard data on disk at that point in time and stores the copy in the specified directory.

Note: backup ignores WAL files and in-memory cache data.

Backups also include a manifest (JSON file) describing what was collected during the backup. Backup filenames reflect the UTC timestamp of when the backup from disk was created, for example:

  • Metastore backup: 20060102T150405Z.meta (includes usernames and passwords)
  • Shard data backup: 20060102T150405Z.<shard_id>.tar.gz
  • Manifest: 20060102T150405Z.manifest

Note: The backup utility copies all data through the meta node used to execute the backup. As a result, performance of a backup and restore is typically limited by the network IO of the meta node. Increasing the resources available to this meta node (such as resizing the EC2 instance) can significantly improve backup and restore performance.

Full versus incremental backup

Backups can be full (using the -full flag) or incremental (by default):

  • Full backups include a copy of the metastore (users, credentials, and permissions) and shard data.
  • Incremental backups include a copy of the metastore and shard data that have changed since the last incremental backup. If there are no existing incremental backups, the system automatically performs a full backup.
  • Keep full backups and incremental backups in separate directories.

Syntax

influxd-ctl [global-options] backup [backup-options] <path-to-backup-directory>

Note: The influxd-ctl backup command exits with 0 for success and 1 for failure. If the backup fails, output can be directed to a log file to troubleshoot.

Global options

For a complete list of the global influxd-ctl options, see the influxd-ctl documentation.

Backup options
  • -db <string>: the name of the single database to back up
  • -from <TCP-address>: the data node TCP address to prefer when backing up
  • -full: perform a full backup
  • -rp <string>: the name of the single retention policy to back up (must specify -db with -rp)
  • -shard <unit>: the ID of the single shard to back up

Backup examples

Store incremental backups in different directories

Store incremental backups in the same directory

Perform an incremental back up

Perform an full back up

Perform an incremental back up on a single database

Restore utility

  • Restores users, roles, databases, and continuous queries.
  • Does not restore Kapacitor subscriptions.
  • Drops data in the new cluster’s _internal database and begins writing to _internal anew.
  • By default, restore writes to databases using the backed-up data’s replication factor. To specify an alternate replication factor when restoring a single database, use the -newrf flag.

Important: Restore backups to a new cluster or existing cluster with a different database name. Otherwise, the restore may fail.

Disable anti-entropy (AE) before restoring a backup

Before restoring a backup, stop the anti-entropy (AE) service (if enabled) on each data node in the cluster, one at a time.

  1. Stop the influxd service.
  2. Set [anti-entropy].enabled to false in the influx configuration file (by default, influx.conf).
  3. Restart the influxd service and wait for the data node to receive read and write requests and for the hinted handoff queue to drain.
  4. Once AE is disabled on all data nodes and each node returns to a healthy state, you’re ready to restore the backup. For details on how to restore your backup, see examples below.
  5. After restoring the backup, restart AE services on each data node.

Syntax

Incremental backup

influxd-ctl [global-options] restore [restore-options] <path-to-backup-directory>

Full backup

influxd-ctl [global-options] restore [options] -full <path-to-manifest-file>
Global options

For a complete list of the global influxd-ctl options, see the influxd-ctl documentation.

Restore options
  • -db <string>: the name of the single database to restore
  • -list: shows the contents of the backup
  • -newdb <string>: the name of the new database to restore to (must specify with -db)
  • -newrf <int>: the new replication factor to restore to (this is capped to the number of data nodes in the cluster)
  • -newrp <string>: the name of the new retention policy to restore to (must specify with -rp)
  • -rp <string>: the name of the single retention policy to restore
  • -shard <unit>: the shard ID to restore

Restore examples

Restore from an incremental backup

Restore from a full backup

Restore from an incremental backup for a single database and give the database a new name

Restore from an incremental backup for a single database and give the database a new name

Common issues with restore

Restore writes information not part of the original backup

Restore a backup created prior to version 1.2.0

Export and import commands

Use the InfluxDB influx_inspect export and influx -import commands to create backup and restore procedures for your disaster recovery and backup strategy. Execute these commands manually or include them in shell scripts to run at scheduled intervals.

Export data

Use the influx_inspect export command to export data in line protocol format from your InfluxDB Enterprise cluster. Options include:

  • Exporting all, or specific, databases
  • Filtering with starting and ending timestamps
  • Using gzip compression for smaller files and faster exports

For details on optional settings and usage, see influx_inspect export command.

In the following example, the database is exported filtered to include only one day and compressed for optimal speed and file size.

influx_inspect export -database myDB -compress -start 2019-05-19T00:00:00.000Z -end 2019-05-19T23:59:59.999Z

Import data

After exporting the data in line protocol format, you can import the data using the influx -import CLI command.

In the following example, the compressed data file is imported into the specified database.

influx -import -database myDB -compress

For details on using the influx -import command, see Import data from a file with -import.

Take AWS snapshots as backup

Use AWS snapshots of data nodes to recover data by exporting line protocol of historical data from shards on disk.

  1. Schedule AWS snapshots. For example, take snapshots of data node directories (include /data directory at minimum, wal directory, and other directories as needed.)
  2. To recover data from a snapshot, create an EC2 system with InfluxDB data node programs (the data node process doesn’t need to run).
  3. Attach the snapshot to your recovery EC2 system. Attach additional volumes as needed for more space.

Note: Extracting shards via influx_inspect (using compress) uses roughly 1.5 times the space as the shard. We recommend provisioning 2.5 times the space that the shards use on disk. (See AWS documentation for procedures to upsize AWS volumes in use.)

  1. Use influx_inspect export to extract line protocol (based on database, rp and time) as needed.
  2. Re-import extracted line protocol.

Run two clusters in separate AWS regions

The advantages to running 2 clusters is you can set the following up in advance:

  • How often to export data from an active cluster to an AWS S3 bucket.
  • How often to replicate the export file in the S3 bucket (use AWS S3 copy command).
  • How often to import data from the AWS S3 bucket to a cluster available for disaster recovery.

To further automate the recovery process, create a customer administration tool to manage users and schedule exports and imports.

First, run two clusters in separate AWS regions, and then transfer your data using custom scripts and S3 buckets.

  1. Create two AWS regions to use:
  • one for disaster recovery (DR)
  • one for actively working with data
  1. In both regions, create separate “availability zones” for each data node in your cluster.

  2. In your “active” region, use influx inspect export to export data from your cluster to an AWS S3 bucket. This S3 bucket automatically replicates cluster data to an S3 bucket in your disaster recovery region.

  3. Create a script to import data (influx-import) from your disaster recovery S3 bucket into your disaster recovery cluster.

  4. (Optional) Create an admin tool to administer data traffic to your clusters and ensure all users go through this tool.

  5. On the “active” cluster, create a script to run on one data node, and then run your script from cron (for example, every 15 minutes) to gather all of the databases. In your script, use a “for loop” to gather data from each database with the influx_inspect tool, and then move data to your S3 bucket.

    Note: Consider your import/export time interval; what is an acceptable time frame difference between your active region and disaster recovery region?

  6. On the “disaster recovery” cluster, create another script to run from cron (for your specified interval). In this script, include the following:

  • Pull files from the S3 bucket and store them locally.

  • Run influx -import to move data to the cluster.

    Note: For best performance, make sure each database has a separate file.

  • Delete files from the S3 bucket.

  1. Run your admin commands, for example, CREATE DB or CREATE USERS on both your disaster recovery and active clusters, which can be handled by your custom admin tool.

  2. Use a separate data node for monitoring metrics.

    Limitations: This solution cannot handle much backfill. To capture backfill, you must create an ad hoc solution. Also, if your disaster recovery site goes down, data on the active site is still getting backed up into S3, but isn’t imported until your disaster recovery cluster is back up and your import script runs.


Set your InfluxDB URL

Upgrade to InfluxDB Cloud or InfluxDB 2.0!

InfluxDB Cloud and InfluxDB OSS 2.0 ready for production.