Backup and Restore

This is archived documentation for InfluxData product versions that are no longer maintained. For newer documentation, see the latest InfluxData documentation.

Backup and Restore in InfluxDB 0.9 are non-functional for a significant subset of users. The underlying storage engine in InfluxDB 0.9 does not lend itself to hot backups. The new storage engine released with InfluxDB 0.10 supports hot incremental backups by design. If reliable hot or incremental backups are an important consideration for you, please upgrade to InfluxDB 0.10 as soon as feasible.

File-level Workaround

If the backup feature does not work on your instance, it is possible to do a file-system backup of InfluxDB.

Backup

  1. Shut down the InfluxDB process
  2. Make a copy of the /data and /meta directories, as well as the configuration file
  3. Restart the InfluxDB process

Note: It is possible to get a working backup without shutting down the process, but some portion of file-level backups against a running database will be non-viable for restore due to race conditions and locking issues during the backup. In our experience at least two out of three backups are viable for servers experiencing ~50k points per second write load. (Query load is largely irrelevant to the race conditions.) Even viable backups may have some data loss for the period immediately preceeding and during the backup process.

Here is an example set of commands:

service influxd stop 
tar -cvf influxdb.backup.2016.01.02.tar /var/opt/influxdb/meta /var/opt/influxdb/data /etc/influxdb/influxdb.conf
...
service influxd start

Note: We recommend using tar without compression to reduce process downtime during backups, or reduce the risk of race conditions if the daemon is still running.

Restore

  1. Shut down the InfluxDB process
  2. Remove the current contents of the /data and /meta directories
  3. Restore the directory contents from the backup
  4. Restart the InfluxDB process

Usage

While a data node is running, you can create a hot backup to a snapshot file (e.g. /tmp/mysnapshot):

$ influxd backup /tmp/mysnapshot

By default, this can only be run from the data node itself. See configuration options below to snapshot from another machine.

To capture a backup from a remote node, specify the host and port using the -host configuration switch:

$ influxd backup -host 10.0.0.0:8088 /tmp/mysnapshot

Once you have your snapshot file, you can copy it to another machine and restore it. Be sure to first shut down any running influxd process

$ influxd restore -config /path/to/influxdb.conf /path/to/mysnapshot

This command will remove the broker and data directories listed in the configuration file provided and replace them with the data in the snapshot. Once the restore is complete make sure the newly written files are readable and writeable by the influxdb user. Once that’s ensured you can start the influxd server normally.

Configuration Options

A configuration section has been added for the snapshot handler with the following defaults:

[snapshot]
enabled = true # Disabled by default if not set.

If set to false InfluxDB will not allow snapshots. Any attempt will return a 404 error. The process must be restarted for configuration changes to take effect.

Implementation

The snapshot file is one or more tar archives that contain a manifest file at the beginning, the data node’s meta file next, and then a list of all shard files. The metastore and shards all use Bolt so they contain a point-in-time copy of the database when the backup was initiated.

The broker node is not backed up because it can be materialized from the data in the data node. The restore command generates a broker meta store based on the highest index in the data node and generates a raft configuration based on the InfluxDB config passed in.

Incremental Backups

InfluxDB also has support for incremental backups. Snapshotting from the server now creates a full backup if one does not exist and creates numbered incremental backups after that.

For example, if you ran:

$ influxd backup /tmp/snapshot

Then you’ll see a full snapshot in /tmp/snapshot. If you run the backup command again using the identical filename then an incremental snapshot will be created at /tmp/snapshot.0. Running it again will create /tmp/snapshot.1, etc.

Running the backup command with a new filename will create a new full backup with that filename, not an incremental backup.

Caveats

This approach currently only works in clusters where the replication factor is the same as the number of nodes in the cluster. A cluster wide backup and restore will be done in the future.