Documentation Home
MySQL 8.0 Reference Manual
Related Documentation Download this Manual
PDF (US Ltr) - 41.6Mb
PDF (A4) - 41.7Mb
HTML Download (TGZ) - 9.3Mb
HTML Download (Zip) - 9.4Mb
Man Pages (TGZ) - 262.1Kb
Man Pages (Zip) - 372.2Kb
Info (Gzip) - 4.0Mb
Info (Zip) - 4.0Mb
Excerpts from this Manual

MySQL 8.0 Reference Manual  /  ...  /  Configuring Audit Logging Characteristics

6.4.5.5 Configuring Audit Logging Characteristics

This section describes how to configure audit logging characteristics, such as the file to which the audit log plugin writes events, the format of written events, whether to enable log file compression and encryption, and space management.

Note

Encryption capabilities described here apply as of MySQL 8.0.17, with the exception of the section that compares current encryption capabilities to the previous more-limited capabilities; see Audit Log File Encryption Prior to MySQL 8.0.17.

For additional information about the functions and system variables that affect audit logging, see Audit Log Functions, and Audit Log Options and Variables.

The audit log plugin can also control which audited events are written to the audit log file, based on event content or the account from which events originate. See Section 6.4.5.7, “Audit Log Filtering”.

Naming Conventions for Audit Log Files

To configure the audit log file name, set the audit_log_file system variable at server startup. The default name is audit.log in the server data directory. For best security, write the audit log to a directory accessible only to the MySQL server and to users with a legitimate reason to view the log.

The plugin interprets the audit_log_file value as composed of an optional leading directory name, a base name, and an optional suffix. If compression or encryption are enabled, the effective file name (the name actually used to create the log file) differs from the configured file name because it has additional suffixes:

  • If compression is enabled, the plugin adds a suffix of .gz.

  • If encryption is enabled, the plugin adds a suffix of .pwd_id.enc, where pwd_id indicates which encryption password to use for log file operations. The audit log plugin stores encryption passwords in the keyring; see Encrypting Audit Log Files.

The effective audit log file name is the name resulting from the addition of applicable compression and encryption suffixes to the configured file name. For example, if the configured audit_log_file value is audit.log, the effective file name is one of the values shown in the following table.

Enabled Features Effective File Name
No compression or encryption audit.log
Compression audit.log.gz
Encryption audit.log.pwd_id.enc
Compression, encryption audit.log.gz.pwd_id.enc

pwd_id indicates the ID of the password used to encrypt or decrypt a file. pwd_id format is pwd_timestamp-seq, where:

  • pwd_timestamp is a UTC value in YYYYMMDDThhmmss format indicating when the password was created.

  • seq is a sequence number. Sequence numbers start at 1 and increase for passwords that have the same pwd_timestamp value.

Here are some example pwd_id password ID values:

20190403T142359-1
20190403T142400-1
20190403T142400-2

To construct the corresponding keyring IDs for storing passwords in the keyring, the audit log plugin adds a prefix of audit_log- to the pwd_id values. For the example password IDs just shown, the corresponding keyring IDs are:

audit_log-20190403T142359-1
audit_log-20190403T142400-1
audit_log-20190403T142400-2

The ID of the password currently used for encryption by the audit log plugin is the one having the largest pwd_timestamp value. If multiple passwords have that pwd_timestamp value, the current password ID is the one with the largest sequence number. For example, in the preceding set of password IDs, two of them have the largest timestamp, 20190403T142400, so the current password ID is the one with the largest sequence number (2).

The audit log plugin performs certain actions during initialization and termination based on the effective audit log file name:

  • During initialization, the plugin checks whether a file with the audit log file name already exists and renames it if so. (In this case, the plugin assumes that the previous server invocation exited unexpectedly with the audit log plugin running.) The plugin then writes to a new empty audit log file.

  • During termination, the plugin renames the audit log file.

  • File renaming (whether during plugin initialization or termination) occurs according to the usual rules for automatic size-based log file rotation; see Manual Audit Log File Rotation.

Selecting Audit Log File Format

To configure the audit log file format, set the audit_log_format system variable at server startup. These formats are available:

  • NEW: New-style XML format. This is the default.

  • OLD: Old-style XML format.

  • JSON: JSON format.

For details about each format, see Section 6.4.5.4, “Audit Log File Formats”.

If you change audit_log_format, it is recommended that you also change audit_log_file. For example, if you set audit_log_format to JSON, set audit_log_file to audit.json. Otherwise, newer log files will have a different format than older files, but they will all have the same base name with nothing to indicate when the format changed.

Compressing Audit Log Files

Audit log file compression can be enabled for any logging format.

To configure audit log file compression, set the audit_log_compression system variable at server startup. Permitted values are NONE (no compression; the default) and GZIP (GNU Zip compression).

If both compression and encryption are enabled, compression occurs before encryption. To recover the original file manually, first decrypt it, then uncompress it. See Manually Uncompressing and Decrypting Audit Log Files.

Encrypting Audit Log Files

Audit log file encryption can be enabled for any logging format. Encryption is based on user-defined passwords (with the exception of the initial password that the audit log plugin generates). To use this feature, the MySQL keyring must be enabled because audit logging uses it for password storage. Any keyring component or plugin can be used; for instructions, see Section 6.4.4, “The MySQL Keyring”.

To configure audit log file encryption, set the audit_log_encryption system variable at server startup. Permitted values are NONE (no encryption; the default) and AES (AES-256-CBC cipher encryption).

To set or get an encryption password at runtime, use these audit log functions:

  • To set the current encryption password, invoke audit_log_encryption_password_set(). This function stores the new password in the keyring. If encryption is enabled, it also performs a log file rotation operation that renames the current log file, and begins a new log file encrypted with the password. File renaming occurs according to the usual rules for automatic size-based log file rotation; see Manual Audit Log File Rotation.

    If the audit_log_password_history_keep_days system variable is nonzero, invoking audit_log_encryption_password_set() also causes expiration of old archived audit log encryption passwords. For information about audit log password history, including password archiving and expiration, see the description of that variable.

  • To get the current encryption password, invoke audit_log_encryption_password_get() with no argument. To get a password by ID, pass an argument that specifies the keyring ID of the current password or an archived password.

    To determine which audit log keyring IDs exist, query the Performance Schema keyring_keys table:

    mysql> SELECT KEY_ID FROM performance_schema.keyring_keys
           WHERE KEY_ID LIKE 'audit_log%'
           ORDER BY KEY_ID;
    +-----------------------------+
    | KEY_ID                      |
    +-----------------------------+
    | audit_log-20190415T152248-1 |
    | audit_log-20190415T153507-1 |
    | audit_log-20190416T125122-1 |
    | audit_log-20190416T141608-1 |
    +-----------------------------+

For additional information about audit log encryption functions, see Audit Log Functions.

When the audit log plugin initializes, if it finds that log file encryption is enabled, it checks whether the keyring contains an audit log encryption password. If not, the plugin automatically generates a random initial encryption password and stores it in the keyring. To discover this password, invoke audit_log_encryption_password_get().

If both compression and encryption are enabled, compression occurs before encryption. To recover the original file manually, first decrypt it, then uncompress it. See Manually Uncompressing and Decrypting Audit Log Files.

Manually Uncompressing and Decrypting Audit Log Files

Audit log files can be uncompressed and decrypted using standard tools. This should be done only for log files that have been closed (archived) and are no longer in use, not for the log file that the audit log plugin is currently writing. You can recognize archived log files because they have been renamed by the audit log plugin to include a timestamp in the file name just after the base name.

For this discussion, assume that audit_log_file is set to audit.log. In that case, an archived audit log file has one of the names shown in the following table.

Enabled Features Archived File Name
No compression or encryption audit.timestamp.log
Compression audit.timestamp.log.gz
Encryption audit.timestamp.log.pwd_id.enc
Compression, encryption audit.timestamp.log.gz.pwd_id.enc

As discussed in Naming Conventions for Audit Log Files, pwd_id format is pwd_timestamp-seq. Thus, the names of archived encrypted log files actually contain two timestamps. The first indicates file rotation time, and the second indicates when the encryption password was created.

Consider the following set of archived encrypted log file names:

audit.20190410T205827.log.20190403T185337-1.enc
audit.20190410T210243.log.20190403T185337-1.enc
audit.20190415T145309.log.20190414T223342-1.enc
audit.20190415T151322.log.20190414T223342-2.enc

Each file name has a unique rotation-time timestamp. By contrast, the password timestamps are not unique:

  • The first two files have the same password ID and sequence number (20190403T185337-1). They have the same encryption password.

  • The second two files have the same password ID (20190414T223342) but different sequence numbers (1, 2). These files have different encryption passwords.

To uncompress a compressed log file manually, use gunzip, gzip -d, or equivalent command. For example:

gunzip -c audit.timestamp.log.gz > audit.timestamp.log

To decrypt an encrypted log file manually, use the openssl command. For example:

openssl enc -d -aes-256-cbc -pass pass:password -md sha256
    -in audit.timestamp.log.pwd_id.enc
    -out audit.timestamp.log

To execute that command, you must obtain password, the encryption password. To do this, use audit_log_encryption_password_get(). For example, if the audit log file name is audit.20190415T151322.log.20190414T223342-2.enc, the password ID is 20190414T223342-2 and the keyring ID is audit-log-20190414T223342-2. Retrieve the keyring password like this:

SELECT audit_log_encryption_password_get('audit-log-20190414T223342-2');

If both compression and encryption are enabled for audit logging, compression occurs before encryption. In this case, the file name has .gz and .pwd_id.enc suffixes added, corresponding to the order in which those operations occur. To recover the original file manually, perform the operations in reverse. That is, first decrypt the file, then uncompress it:

openssl enc -d -aes-256-cbc -pass pass:password -md sha256
    -in audit.timestamp.log.gz.pwd_id.enc
    -out audit.timestamp.log.gz
gunzip -c audit.timestamp.log.gz > audit.timestamp.log
Audit Log File Encryption Prior to MySQL 8.0.17

This section covers the differences in audit log file encryption capabilities prior to and as of MySQL 8.0.17, which is when password history was implemented (which includes password archiving and expiration). It also indicates how the audit log plugin handles upgrades to MySQL 8.0.17 or higher from versions lower than 8.0.17.

Feature Prior to MySQL 8.0.17 As of MySQL 8.0.17
Number of passwords Single password only Multiple passwords permitted
Encrypted log file names .enc suffix .pwd_id.enc suffix
Password keyring ID audit_log audit_log-pwd_id
Password history No Yes

Prior to MySQL 8.0.17, there is no password history, so setting a new password makes the old password inaccessible, rendering MySQL Enterprise Audit unable to read log files encrypted with the old password. Should you anticipate a need to decrypt those files manually, you must maintain a record of previous passwords.

If audit log file encryption is enabled when you upgrade to MySQL 8.0.17 or higher from a lower version, the audit log plugin performs these upgrade actions:

  • During plugin initialization, the plugin checks for an encryption password with a keyring ID of audit_log. If it finds one, the plugin duplicates the password using a keyring ID in audit_log-pwd_id format and uses it as the current encryption password. (For details about pwd_id syntax, see Naming Conventions for Audit Log Files.)

  • Existing encrypted log files have a suffix of .enc. The plugin does not rename these to have a suffix of .pwd_id.enc, but can read them as long as the key with the ID of audit_log remains in the keyring.

  • When password cleanup occurs, if the plugin expires any password with a keyring ID in audit_log-pwd_id format, it also expires the password with a keyring ID of audit_log, if it exists. (At this point, encrypted log files that have a suffix of .enc rather than .pwd_id.enc become unreadable by the plugin, so it is assumed that you no longer need them.)

Space Management of Audit Log Files

The audit log file has the potential to grow quite large and consume a great deal of disk space. To manage the space used, employ these methods:

  • Log file rotation. This involves rotating the current log file by renaming it, then opening a new current log file using the original name. Rotation can be performed manually, or configured to occur automatically.

  • Pruning of rotated JSON-format log files, if automatic rotation is enabled. Pruning can be performed based on log file age (as of MySQL 8.0.24), or combined log file size (as of MySQL 8.0.26).

To configure audit log file space management, use the following system variables:

Note

For JSON-format log files, rotation also occurs when the value of the audit_log_format_unix_timestamp system variable is changed at runtime. However, this does not occur for space-management purposes, but rather so that, for a given JSON-format log file, all records in the file either do or do not include the time field.

Note

Rotated (renamed) log files are not removed automatically. For example, with size-based log file rotation, renamed log files have unique names and accumulate indefinitely. They do not rotate off the end of the name sequence. To avoid excessive use of space:

  • As of MySQL 8.0.24 (for JSON-format log files): Enable log file pruning as described in Audit Log File Pruning.

  • Otherwise (for non-JSON files, or prior to MySQL 8.0.24 for all log formats): Remove old files periodically, backing them up first as necessary. If backed-up log files are encrypted, also back up the corresponding encryption passwords to a safe place, should you need to decrypt the files later.

The following sections describe log file rotation and pruning in greater detail.

Manual Audit Log File Rotation

If audit_log_rotate_on_size is 0 (the default), no log rotation occurs unless performed manually. In this case, the audit log plugin closes and reopens the log file when the audit_log_flush value changes from disabled to enabled. Log file renaming must be done externally to the server. Suppose that the log file name is audit.log and you want to maintain the three most recent log files, cycling through the names audit.log.1 through audit.log.3. On Unix, perform rotation manually like this:

  1. From the command line, rename the current log files:

    mv audit.log.2 audit.log.3
    mv audit.log.1 audit.log.2
    mv audit.log audit.log.1

    This strategy overwrites the current audit.log.3 contents, placing a bound on the number of archived log files and the space they use.

  2. At this point, the plugin is still writing to the current log file, which has been renamed to audit.log.1. Connect to the server and flush the log file so the plugin closes it and reopens a new audit.log file:

    SET GLOBAL audit_log_flush = ON;

    audit_log_flush is special in that its value remains OFF so that you need not disable it explicitly before enabling it again to perform another flush.

Note

If compression or encryption are enabled, log file names include suffixes that signify the enabled features, as well as a password ID if encryption is enabled. If file names include a password ID, be sure to retain the ID in the name of any files you rename manually so that the password to use for decryption operations can be determined.

Note

For JSON-format logging, renaming audit log files manually makes them unavailable to the log-reading functions because the audit log plugin can no longer determine that they are part of the log file sequence (see Section 6.4.5.6, “Reading Audit Log Files”). Consider setting audit_log_rotate_on_size greater than 0 to use size-based rotation instead.

Automatic Audit Log File Rotation

If audit_log_rotate_on_size is greater than 0, setting audit_log_flush has no effect. Instead, whenever a write to the current log file causes its size to exceed the audit_log_rotate_on_size value, the audit log plugin automatically renames the current log file and opens a new current log file using the original name.

Automatic size-based rotation also occurs under these conditions:

The plugin renames the original file by inserting a timestamp just after its base name. For example, if the file name is audit.log, the plugin renames it to a value such as audit.20210115T140633.log. The timestamp is a UTC value in YYYYMMDDThhmmss format. For XML logging, the timestamp indicates rotation time. For JSON logging, the timestamp is that of the last event written to the file.

If log files are encrypted, the original file name already contains a timestamp indicating the encryption password creation time (see Naming Conventions for Audit Log Files). In this case, the file name after rotation contains two timestamps. For example, an encrypted log file named audit.log.20210110T130749-1.enc is renamed to a value such as audit.20210115T140633.log.20210110T130749-1.enc.

Audit Log File Pruning

The audit log plugin supports pruning of rotated JSON-format audit log files, if automatic log file rotation is enabled. To use this capability:

  • Set audit_log_format to JSON. (In addition, consider also changing audit_log_file; see Selecting Audit Log File Format.)

  • Set audit_log_rotate_on_size greater than 0 to specify the size in bytes at which automatic log file rotation occurs.

  • By default, no pruning of automatically rotated JSON-format log files occurs. To enable pruning, set one of these system variables to a value greater than 0:

    Nonzero values of audit_log_max_size take precedence over nonzero values of audit_log_prune_seconds. If both are set greater than 0 at plugin initialization, a warning is written to the server error log. If a client sets both greater than 0 at runtime, a warning is returned to the client.

    Note

    Warnings to the error log are written as Notes, which are information messages. To ensure that such messages appear in the error log and are not discarded, make sure that error-logging verbosity is sufficient to include information messages. For example, if you are using priority-based log filtering, as described in Section 5.4.2.5, “Priority-Based Error Log Filtering (log_filter_internal)”, set the log_error_verbosity system variable to a value of 3.

Pruning of JSON-format log files, if enabled, occurs as follows:

For pruning based on combined rotated log file size, if the combined size is greater than the limit specified by audit_log_max_size, the audit log plugin removes the oldest files until their combined size does not exceed the limit.

For pruning based on rotated log file age, the pruning point is the current time minus the value of audit_log_prune_seconds. In rotated JSON-format log files, the timestamp part of each file name indicates the timestamp of the last event written to the file. The audit log plugin uses file name timestamps to determine which files contain only events older than the pruning point, and removes them.

Write Strategies for Audit Logging

The audit log plugin can use any of several strategies for log writes. Regardless of strategy, logging occurs on a best-effort basis, with no guarantee of consistency.

To specify a write strategy, set the audit_log_strategy system variable at server startup. By default, the strategy value is ASYNCHRONOUS and the plugin logs asynchronously to a buffer, waiting if the buffer is full. It's possible to tell the plugin not to wait (PERFORMANCE) or to log synchronously, either using file system caching (SEMISYNCHRONOUS) or forcing output with a sync() call after each write request (SYNCHRONOUS).

For asynchronous write strategy, the audit_log_buffer_size system variable is the buffer size in bytes. Set this variable at server startup to change the buffer size. The plugin uses a single buffer, which it allocates when it initializes and removes when it terminates. The plugin does not allocate this buffer for nonasynchronous write strategies.

Asynchronous logging strategy has these characteristics:

  • Minimal impact on server performance and scalability.

  • Blocking of threads that generate audit events for the shortest possible time; that is, time to allocate the buffer plus time to copy the event to the buffer.

  • Output goes to the buffer. A separate thread handles writes from the buffer to the log file.

With asynchronous logging, the integrity of the log file may be compromised if a problem occurs during a write to the file or if the plugin does not shut down cleanly (for example, in the event that the server host exits unexpectedly). To reduce this risk, set audit_log_strategy to use synchronous logging.

A disadvantage of PERFORMANCE strategy is that it drops events when the buffer is full. For a heavily loaded server, the audit log may have events missing.