open-vault/website/source/api/index.html.md

206 lines
7.1 KiB
Markdown
Raw Normal View History

2015-04-10 05:52:02 +00:00
---
2017-03-17 18:06:03 +00:00
layout: "api"
2015-04-10 05:52:02 +00:00
page_title: "HTTP API"
2015-04-18 21:34:47 +00:00
sidebar_current: "docs-http-overview"
2015-04-10 05:52:02 +00:00
description: |-
Vault has an HTTP API that can be used to control every aspect of Vault.
---
# HTTP API
The Vault HTTP API gives you full access to Vault via HTTP. Every aspect of
Vault can be controlled via this API. The Vault CLI uses the HTTP API to access
Vault.
2015-04-10 05:52:02 +00:00
All API routes are prefixed with `/v1/`.
This documentation is only for the v1 API, which is currently the only version.
2015-04-10 05:52:02 +00:00
~> **Backwards compatibility:** At the current version, Vault does not yet
promise backwards compatibility even with the v1 prefix. We'll remove this
warning when this policy changes. At this point in time the core API (that
is, `sys/` routes) change very infrequently, but various secrets engines/auth
methods/etc. sometimes have minor changes to accommodate new features as
they're developed.
2015-04-10 05:52:02 +00:00
## Transport
The API is expected to be accessed over a TLS connection at all times, with a
valid certificate that is verified by a well-behaved client. It is possible to
disable TLS verification for listeners, however, so API clients should expect
to have to do both depending on user settings.
2015-04-10 05:52:02 +00:00
## Authentication
Once Vault is unsealed, almost every other operation requires a _client token_.
A user may have a client token sent to her. The client token must be sent as
the `X-Vault-Token` HTTP header.
2015-04-10 05:52:02 +00:00
Otherwise, a client token can be retrieved via [authentication
backends](/docs/auth/index.html).
2015-04-10 05:52:02 +00:00
Each auth method has one or more unauthenticated login endpoints. These
endpoints can be reached without any authentication, and are used for
authentication to Vault itself. These endpoints are specific to each auth
method.
Responses from auth login methods that generate an authentication token are
sent back to the client via JSON. The resulting token should be saved on the
client or passed via the `X-Vault-Token` header for future requests.
## Namespaces
2015-04-10 05:52:02 +00:00
If using the [Namespaces](/docs/enterprise/namespaces/index.html) feature, API
operations are relative to the namespace value passed in via the
`X-Vault-Namespace` header. For instance, if the request path is to
`secret/foo`, and the header is set to `ns1/ns2/`, the final request path Vault
uses will be `ns1/ns2/secret/foo`. Note that it is semantically equivalent to
use a full path rather than the `X-Vault-Namespace` header, as the operation in
Vault will always look up the correct namespace based on the final given path.
Thus, it would be equivalent to the above example to set `X-Vault-Namespace` to
`ns1/` and a request path of `ns2/secret/foo`, or to not set
`X-Vault-Namespace` at all and use a request path of `ns1/ns2/secret/foo`.
2015-04-10 05:52:02 +00:00
## API Operations
With few documented exceptions, all request body data and response data from
Vault is via JSON. Vault will set the `Content-Type` header appropriately but
does not require that clients set it.
Different plugins implement different APIs according to their functionality.
The examples below are created with the `KVv1` backend, which acts like a very
simple Key/Value store. Read the documentation for a particular backend for
detailed information on its API; this simply provides a general overview.
For `KVv1`, reading a secret via the HTTP API is done by issuing a GET:
2015-06-01 04:23:44 +00:00
```text
/v1/secret/foo
```
This maps to `secret/foo` where `foo` is the key in the `secret/` mount, which
is mounted by default on a fresh Vault install and is of type `kv`.
Here is an example of reading a secret using cURL:
2015-06-01 04:23:44 +00:00
```shell
$ curl \
-H "X-Vault-Token: f3b09679-3001-009d-2b80-9c306ab81aa6" \
-X GET \
http://127.0.0.1:8200/v1/secret/foo
2015-06-01 04:23:44 +00:00
```
A few endpoints consume query parameters via `GET` calls, but only if those
parameters are not sensitive, as some load balancers will log these. Most
endpoints that consume parameters use `POST` instead and put the parameters in
the request body.
You can list secrets as well. To do this, either issue a GET with the query
parameter `list=true`, or you can use the `LIST` HTTP verb. For the `kv`
backend, listing is allowed on directories only, and returns the keys in the
given directory:
```shell
$ curl \
-H "X-Vault-Token: f3b09679-3001-009d-2b80-9c306ab81aa6" \
-X LIST \
http://127.0.0.1:8200/v1/secret/
```
The API documentation uses `LIST` as the HTTP verb, but you can still use `GET`
with the `?list=true` query string.
To use an API that consumes data via request body, issue a `POST` or `PUT`:
2015-06-01 04:23:44 +00:00
```text
/v1/secret/foo
```
2015-06-01 04:23:44 +00:00
with a JSON body like:
```javascript
{
"value": "bar"
}
```
Here is an example of writing a secret using cURL:
2015-06-01 04:23:44 +00:00
```shell
$ curl \
-H "X-Vault-Token: f3b09679-3001-009d-2b80-9c306ab81aa6" \
-H "Content-Type: application/json" \
-X POST \
-d '{"value":"bar"}' \
http://127.0.0.1:8200/v1/secret/baz
2015-06-01 04:23:44 +00:00
```
Vault currently considers `PUT` and `POST` to be synonyms. Rather than trust a
client's stated intentions, Vault backends can implement an existence check to
discover whether an operation is actually a create or update operation based on
the data already stored within Vault. This makes permission management via ACLs
more flexible.
2015-06-01 04:23:44 +00:00
For more examples, please look at the Vault API client.
2015-04-10 05:52:02 +00:00
## Help
To retrieve the help for any API within Vault, including mounted backends, auth
methods, etc. then append `?help=1` to any URL. If you have valid permission to
access the path, then the help text will be returned with the following
structure:
2015-04-10 05:52:02 +00:00
```javascript
{
"help": "help text"
}
```
2015-04-10 05:52:02 +00:00
## Error Response
A common JSON structure is always returned to return errors:
```javascript
{
"errors": [
"message",
"another message"
]
}
```
2015-04-10 05:52:02 +00:00
This structure will be sent down for any HTTP status greater than
or equal to 400.
## HTTP Status Codes
The following HTTP status codes are used throughout the API. Vault tries to
adhere to these whenever possible, but in some cases may not -- feel free to
file a bug in that case to point our attention to it!
2015-04-10 05:52:02 +00:00
- `200` - Success with data.
- `204` - Success, no data returned.
- `400` - Invalid request, missing or invalid data.
- `403` - Forbidden, your authentication details are either incorrect, you
don't have access to this feature, or - if CORS is enabled - you made a
cross-origin request from an origin that is not allowed to make such
requests.
- `404` - Invalid path. This can both mean that the path truly doesn't exist or
that you don't have permission to view a specific path. We use 404 in some
cases to avoid state leakage.
- `429` - Default return code for health status of standby nodes. This will
likely change in the future.
- `473` - Default return code for health status of performance standby nodes.
- `500` - Internal server error. An internal error has occurred, try again
later. If the error persists, report a bug.
- `502` - A request to Vault required Vault making a request to a third party;
the third party responded with an error of some kind.
- `503` - Vault is down for maintenance or is currently sealed. Try again
later.
## Limits
A maximum request size of 32MB is imposed to prevent a denial of service attack
with arbitrarily large requests; this can be tuned per `listener` block in
Vault's server configuration file.