The custom_fields_by_lua configuration allows for the dynamic modification of
log fields using Lua code. Below is a snippet of an example configuration that
removes the route field from the logs:
curl -i -X POST http://localhost:8001/plugins \
...
--data config.custom_fields_by_lua.route="return nil"
Similarly, new fields can be added:
curl -i -X POST http://localhost:8001/plugins \
...
--data config.custom_fields_by_lua.header="return kong.request.get_header('h1')"
Dot characters (.) in the field key create nested fields. You can use a backslash \ to escape a dot if you want to keep it in the field name.
For example, if you configure a field in the File Log plugin with both a regular dot and an escaped dot:
curl -i -X POST http://localhost:8001/plugins/ \
...
--data config.name=file-log \
--data config.custom_fields_by_lua[my_file.log\.field]="return foo"
The field will look like this in the log:
"my_file": {
"log.field": "foo"
}
All logging plugins use the same table for logging.
If you set custom_fields_by_lua in one plugin, all logging plugins that execute after that plugin will also use the same configuration.
For example, if you configure fields via custom_fields_by_lua in File Log, those same fields will appear in Kafka Log, since File Log executes first.
If you want all logging plugins to use the same configuration, we recommend using the Pre-function plugin to call kong.log.set_serialize_value so that the function is applied predictably and is easier to manage.
If you don’t want all logging plugins to use the same configuration, you need to manually disable the relevant fields in each plugin.
For example, if you configure a field in File Log that you don’t want appearing in Kafka Log, set that field to return nil in the Kafka Log plugin:
curl -i -X POST http://localhost:8001/plugins/ \
...
--data config.name=kafka-log \
--data config.custom_fields_by_lua.my_file_log_field="return nil"
See the plugin execution order reference for more details on plugin ordering.
Lua code runs in a restricted sandbox environment, whose behavior is governed
by the untrusted_lua configuration properties.
Sandboxing consists of several limitations in the way the Lua code can be executed,
for heightened security.
The following functions are not available because they can be used to abuse the system:
-
string.rep: Can be used to allocate millions of bytes in one operation.
-
{set|get}metatable: Can be used to modify the metatables of global objects (strings, numbers).
-
collectgarbage: Can be abused to kill the performance of other workers.
-
_G: Is the root node which has access to all functions. It is masked by a temporary table.
-
load{file|string}: Is deemed unsafe because it can grant access to the global environment.
-
raw{get|set|equal}: Potentially unsafe because sandboxing relies on some metatable manipulation.
-
string.dump: Can display confidential server information (such as implementation of functions).
-
math.randomseed: Can affect the host system. Kong Gateway already seeds the random number generator properly.
- All
os.* (except os.clock, os.difftime, and os.time). os.execute can significantly alter the host system.
-
io.*: Provides access to the hard drive.
-
dofile|require: Provides access to the hard drive.
The exclusion of require means that plugins must only use PDK functions kong.*. The ngx.* abstraction is
also available, but it is not guaranteed to be present in future versions of the plugin.
In addition to the above restrictions:
- All the provided modules (like
string or table) are read-only and can’t be modified.
- Bytecode execution is disabled.
- The
kong.cache points to a cache instance that is dedicated to the Serverless Functions plugins. It does not provide access to the global Kong Gateway cache. It only exposes the get method. Explicit write operations like set or invalidate are not available.
Further, as code runs in the context of the log phase, only PDK methods
that can run in said phase can be used.
