ArangoDB v3.13 is under development and not released yet. This documentation is not final and potentially incomplete.

HTTP interface for Hot Backups

The HTTP API for Hot Backups lets you manage incremental, zero-downtime data backups

ArangoDB Enterprise Edition

Hot Backups are near instantaneous consistent snapshots of an entire ArangoDB deployment. This includes all databases, collections, indexes, Views, graphs, and users at any given time.

For creations a label may be specified, which if omitted is replaced with a generated UUID. This label is then combined with a timestamp to generate an identifier for the created hot backup. Subsequently, all other APIs operate on these identifiers.

Make sure to understand all aspects of Hot Backups, most of all the requirements and limitations, before using the API.

The permissions required to use the /_admin/backup/* endpoints depends on the setting of the --backup.options-api startup option.

Create a backup

POST /_admin/backup/create
Creates a consistent local backup “as soon as possible”, very much like a snapshot in time, with a given label. The ambiguity in the phrase “as soon as possible” refers to the next window during which a global write lock across all databases can be obtained in order to guarantee consistency. Note that the backup at first resides on the same machine and hard drive as the original data. Make sure to upload it to a remote site for an actual backup.
Request Body application/json object
  • If this flag is set to true and no global transaction lock can be acquired within the given timeout, a possibly inconsistent backup is taken. The default for this flag is false and in this case a timeout results in an HTTP 408 error.

  • If this flag is set to true and no global transaction lock can be acquired within the given timeout, all running transactions are forcefully aborted to ensure that a consistent backup can be created. This does not include JavaScript transactions. It waits for the transactions to be aborted at most timeout seconds. Thus using force the request timeout is doubled. To abort transactions is almost certainly not what you want for your application. In the presence of intermediate commits it can even destroy the atomicity of your transactions. Use at your own risk, and only if you need a consistent backup at all costs. The default and recommended value is false. If both allowInconsistent and force are set to true, then the latter takes precedence and transactions are aborted. This is only available in the cluster.

  • The label for this backup. The label is used together with a timestamp string create a unique backup identifier, <timestamp>_<label>. If no label is specified, the empty string is assumed and a default UUID is created for this part of the ID.

  • The time in seconds that the operation tries to get a consistent snapshot. The default is 120 seconds.

Responses
  • If all is well, code 201 is returned.

  • If the create command is invoked with bad parameters or any HTTP method other than POST, then an HTTP 400 is returned. The specifics are detailed in the returned error document.

  • If the operation cannot obtain a global transaction lock within the timeout, then an HTTP 408 is returned.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/create
{
  "label": "foo"
}
Show output

Restore a backup

POST /_admin/backup/restore
Restores a consistent local backup from a snapshot in time, with a given id. The backup snapshot must reside on the ArangoDB service locally.
Request Body application/json object
  • The id of the backup to restore from.

Responses
  • Is returned if the backup could be restored. Note that there is an inevitable discrepancy between the single server and the cluster. In a single server, the request returns successfully, but the restore is only executed afterwards. In the cluster, the request only returns when the restore operation has been completed successfully. The cluster behaviour is obviously the desired one, but in a single instance, one cannot keep a connection open across a restart.

  • If the restore command is invoked with bad parameters or any HTTP method other than POST, then an HTTP 400 is returned. The specifics are detailed in the returned error document.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/restore
{
  "id": "2024-03-11T19.03.42Z_538c5497-3fe1-4df3-9284-b485c3c26245"
}
Show output

Delete a backup

POST /_admin/backup/delete
Delete a specific local backup identified by the given id.
Request Body application/json object
  • The identifier for this backup.

Responses
  • If all is well, this code 200 is returned.

  • If the delete command is invoked with bad parameters or any HTTP method other than POST, then an HTTP 400 is returned.

  • If a backup corresponding to the identifier id cannot be found.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/delete
{
  "id": "2024-03-11T19.03.45Z_0bbb3d5b-567a-4ea8-b3ca-16006746f7ac"
}
Show output

List all backups

POST /_admin/backup/list
Lists all locally found backups.
Request Body application/json object
  • The body can either be empty (in which case all available backups are listed), or it can be an object with an attribute id, which is a string. In the latter case the returned list is restricted to the backup with the given id.

Responses
  • If all is well, code 200 is returned.

  • If the list command is invoked with bad parameters, then an HTTP 400 is returned.

  • If an id or a list of ids was given and the given ids were not found as identifiers of a backup, an HTTP 404 Not Found is returned.

  • If the list command is invoked with any HTTP method other than POST, then an HTTP 405 Method Not Allowed is returned.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/list
{}
Show output

Upload a backup to a remote repository

POST /_admin/backup/upload
Upload a specific local backup to a remote repository, or query progress on a previously scheduled upload operation, or abort a running upload operation.
Request Body application/json object
  • Set this to true if a running upload operation should be aborted. In this case, the only other body parameter which is needed is uploadId.

  • Configuration of remote repository. This is required when an upload operation is scheduled. In this case leave out the uploadId attribute. See the description of the arangobackup program in the manual for a description of the config object.

  • The identifier for this backup. This is required when an upload operation is scheduled. In this case leave out the uploadId attribute.

  • URL of remote repository. This is required when an upload operation is scheduled. In this case leave out the uploadId attribute. Provided repository URLs are normalized and validated as follows: One single colon must appear separating the configuration section name and the path. The URL prefix up to the colon must exist as a key in the config object below. No slashes must appear before the colon. Multiple back to back slashes are collapsed to one, as .. and . are applied accordingly. Local repositories must be absolute paths and must begin with a /. Trailing / are removed.

  • Upload ID to specify for which upload operation progress is queried or the upload operation to abort. If you specify this, leave out all the above body parameters.

Responses
  • If all is well, code 200 is returned if progress is inquired or the operation is aborted.

  • If all is well, code 202 is returned if a new operation is scheduled.

  • If the upload command is invoked with bad parameters or any HTTP method other than POST, then an HTTP 400 is returned.

  • If the authentication to the remote repository fails, then an HTTP 400 is returned.

  • If a backup corresponding to the identifier id cannot be found, or if there is no known upload operation with the given uploadId.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/upload
{
  "id": "2024-03-11T19.03.45Z_eed916f4-46cf-4dce-8cee-f156716a9fec",
  "remoteRepository": "local://tmp/backups",
  "config": {
    "local": {
      "type": "local"
    }
  }
}
Show output

The result object of the body holds the uploadId string attribute which can be used to follow the upload process.

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/upload
{
  "uploadId": "68604"
}
Show output

Download a backup from a remote repository

POST /_admin/backup/download
Download a specific local backup from a remote repository, or query progress on a previously scheduled download operation, or abort a running download operation.
Request Body application/json object
  • Set this to true if a running download operation should be aborted. In this case, the only other body parameter which is needed is downloadId.

  • Configuration of remote repository. This is required when a download operation is scheduled. In this case leave out the downloadId attribute. See the description of the arangobackup program in the manual for a description of the config object.

  • Download ID to specify for which download operation progress is queried, or the download operation to abort. If you specify this, leave out all the above body parameters.

  • The identifier for this backup. This is required when a download operation is scheduled. In this case leave out the downloadId attribute.

  • URL of remote repository. This is required when a download operation is scheduled. In this case leave out the downloadId attribute. Provided repository URLs are normalized and validated as follows: One single colon must appear separating the configuration section name and the path. The URL prefix up to the colon must exist as a key in the config object below. No slashes must appear before the colon. Multiple back to back slashes are collapsed to one, as .. and . are applied accordingly. Local repositories must be absolute paths and must begin with a /. Trailing / are removed.

Responses
  • If all is well, code 200 is returned if progress is inquired or the operation is aborted.

  • If all is well, code 202 is returned if a new operation is scheduled.

  • If the download command is invoked with bad parameters or any HTTP method other than POST, then an HTTP 400 is returned.

  • If the authentication to the remote repository fails, then an HTTP 401 is returned.

  • If a backup corresponding to the identifier id cannot be found, or if there is no known download operation with the given downloadId.

Examples

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/download
{
  "id": "2024-03-11T19.03.45Z_81ef7b20-1787-41a4-956a-1e3835965141",
  "remoteRepository": "local://tmp/backups",
  "config": {
    "local": {
      "type": "local"
    }
  }
}
Show output

The result object of the body holds the downloadId string attribute which can be used to follow the download process.

curl -X POST --header 'accept: application/json' --data-binary @- --dump - http://localhost:8529/_admin/backup/download
{
  "downloadId": "68694"
}
Show output