Audit devices are the components in Vault that collectively keep a detailed log of all requests to Vault, and their responses. Because every operation with Vault is an API request/response, when using a single audit device, the audit log contains every interaction with the Vault API, including errors - except for a few paths which do not go via the audit system.
The non-audited paths are:
and also, if the relevant listener configuration settings allow unauthenticated access:
When multiple audit devices are enabled, Vault will attempt to send the audit logs to all of them. This allows you to not only have redundant copies, but also a way to check for data tampering in the logs themselves.
Vault considers a request to be successful if it can log to at least one configured audit device (see: Blocked Audit Devices section below). Therefore in order to build a complete picture of all audited actions, use the aggregate/union of the logs from each audit device.
Note: It is highly recommended that you configure Vault to use multiple audit devices. Audit failures can prevent Vault from servicing requests, so it is important to provide at least one other device.
Each line in the audit log is a JSON object. The
type field specifies what
type of object it is. Currently, only two types exist:
The line contains all of the information for any given request and response. By
default, all the sensitive information is first hashed before logging in the
The audit logs contain the full request and response objects for every interaction with Vault. The request and response can be matched utilizing a unique identifier assigned to each request.
Most strings contained within requests and responses are hashed with a salt using HMAC-SHA256. The purpose of the hash is so that secrets aren't in plaintext within your audit logs. However, you're still able to check the value of secrets by generating HMACs yourself; this can be done with the audit device's hash function and salt by using the
/sys/audit-hash API endpoint (see the documentation for more details).
Currently, only strings that come from JSON or returned in JSON are HMAC'd. Other data types, like integers, booleans, and so on, are passed through in plaintext. We recommend that all sensitive data be provided as string values inside all JSON sent to Vault (i.e., that integer values are provided in quotes).
While most strings are hashed, Vault does make some exceptions, such as auth and secrets, and users can enable additional exceptions using the secrets enable command, and then tune it afterward.
When a Vault server is first initialized, no auditing is enabled. Audit
devices must be enabled by a root user using
vault audit enable.
When enabling an audit device, options can be passed to it to configure it. For example, the command below enables the file audit device:
$ vault audit enable file file_path=/var/log/vault_audit.log
In the command above, we passed the "file_path" parameter to specify the path where the audit log will be written to. Each audit device has its own set of parameters. See the documentation to the left for more details.
Note: Audit device configuration is replicated to all nodes within a
cluster by default, and to performance/DR secondaries for Vault Enterprise clusters.
Before enabling an audit device, ensure that all nodes within the cluster(s)
will be able to successfully log to the audit device to avoid Vault being
blocked from serving requests.
An audit device can be limited to only within the node's cluster with the
When an audit device is disabled, it will stop receiving logs immediately. The existing logs that it did store are untouched.
Note: Once an audit device is disabled, you will no longer be able to HMAC values for comparison with entries in the audit logs. This is true even if you re-enable the audit device at the same path, as a new salt will be created for hashing.
Audit device logs are critically important and ignoring auditing failures opens an avenue for attack. Vault will not respond to requests when no enabled audit devices can record them.
Vault can distinguish between two types of audit device failures.
A blocking failure is one where an attempt to write to the audit device never completes. This is unlikely with a local disk device, but could occure with a network-based audit device.
When multiple audit devices are enabled, if any of them fail in a non-blocking fashion, Vault requests can still complete successfully provided at least one audit device successfully writes the audit record. If any of the audit devices fail in a blocking fashion however, Vault requests will hang until the blocking is resolved.
In other words, Vault will not complete any requests until the blocked audit device can write.
Refer to Blocked Audit Devices for a step-by-step tutorial.
Audit devices also have a full HTTP API. Please see the Audit device API docs for more details.
(string: "json")- Allows selecting the output format. Valid values are
"jsonx", which formats the normal log entries as XML.
(bool: true)- If enabled, enables the hashing of token accessor.
(bool: false)- If enabled, logs the security sensitive information without hashing, in the raw format.
(string: "")- A customizable string prefix to write before the actual log line.
Some Vault responses can be very large. Primarily, this affects list operations - as Vault lacks pagination in its APIs, listing a very large collection can result in a response that is tens of megabytes long. Some audit backends are unable to process individual audit records of larger sizes.
The contents of the response for a list operation is often not very interesting;
most contain only a "keys" field, containing a list of IDs. Select API endpoints
additionally return a "key_info" field, a map from ID to some additional
information about the list entry -
identity/entity/id/ is an example of this.
Even in this case, the response to a list operation is usually less-confidential
or public information, for which having the full response in the audit logs is of
elide_list_responses audit option provides the flexibility to not write the
full list response data from the audit log, to mitigate the creation of very long
individual audit records.
When enabled, it affects only audit records of
request.operation=list. The values of
response.data.key_info will be replaced with a simple integer, recording how
many entries were contained in the list (
keys) or map (
key_info) - therefore
even with this feature enabled, it is still possible to see how many items were
returned by a list operation.
This extra processing only affects the response data fields
and only when they have the expected data types - in the event a list response
contains data outside of the usual conventions that apply to Vault list responses,
it will be left as is by this feature.
Here is an example of an audit record that has been processed by this feature (formatted with extra whitespace, and with fields not relevant to the example omitted):