Log Configuration File

The validator and the Python SDK make it easy to customize Sawtooth logging output. This is done by creating a log config file in TOML or YAML format and passing it to the built-in Python logging module.

Note

You must use YAML to configure a remote syslog service. Due to a limitation in the TOML spec, you cannot configure a remote syslog service using TOML.

About Sawtooth Log Files

If there is no log configuration file provided, Sawtooth creates an error log and a debug log by default. These files are stored in the log directory (log_dir), which is /var/log/sawtooth by default. However, the location SAWTOOTH_HOME environment variable, if set, can change the default location. For more information, see Path Configuration File.

For Sawtooth core components, such as the validator or REST API, the log file names are {component}-debug.log and {component}-error.log. For example, the validator log files are:

  • validator-debug.log

  • validator-error.log

For Python transaction processors, the author determines the name of the log file. Because a Sawtooth node can run several instances of the same transaction processor, it is important to create unique log file names for each running transaction processor to avoid naming conflicts.

The example transaction processors included in Sawtooth use the following naming convention:

  • {TPname}-{zmqID}-debug.log

  • {TPname}-{zmqID}-error.log

For example, an instance of the IntegerKey (intkey) transaction processor could have the following log files:

  • intkey-18670799cbbe4367-debug.log

  • intkey-18670799cbbe4367-error.log

Log Configuration

To change the default logging behavior of a Sawtooth component, such as the validator or REST API, put a log configuration file in the config directory (/var/log/sawtooth by default; see Path Configuration File).

Sawtooth provides an example log configuration file in /etc/sawtooth/log_config.toml.example. To create a log configuration file, copy the example file to the config directory and name it log_config.toml.

Each transaction processor can define its own config file. The name of this file is determined by the author. The transaction processors included in the Python SDK use the following naming convention:

  • {TransactionFamilyName}_log_config.toml

For example, the IntegerKey (intkey) log configuration file is intkey_log_config.toml.

Example: Configure a Specific Logger

If the default logs give too much information, you can configure a specific logger that will only report on the area of the code you are interested in.

This example log_config.toml file creates a handler that only writes interconnect logs to the directory and file specified.

version = 1
disable_existing_loggers = false

[formatters.simple]
format = "[%(asctime)s.%(msecs)03d [%(threadName)s] %(module)s %(levelname)s] %(message)s"
datefmt = "%H:%M:%S"

[handlers.interconnect]
level = "DEBUG"
formatter = "simple"
class = "logging.FileHandler"
filename = "path/filename.log"

[loggers."sawtooth_validator.networking.interconnect"]
level = "DEBUG"
propagate = true
handlers = [ "interconnect"]

The formatter and log level can also be specified to provide the exact information you want in your logs.

Example: Create a Rotating File Handler

This example shows how to set up rotating logs. This is useful when the logs may grow very large, such as with a long-running network. For example:

 [formatters.simple]
 format = "[%(asctime)s.%(msecs)03d [%(threadName)s] %(module)s %(levelname)s] %(message)s"
 datefmt = "%H:%M:%S"

 [handlers.interconnect]
 level = "DEBUG"
 formatter = "simple"
 class = "logging.handlers.RotatingFileHandler"
 filename = "example-interconnect.log"
 maxBytes = 50000000
 backupCount=20

[loggers."sawtooth_validator.networking.interconnect"]
 level = "DEBUG"
 propagate = true
 handlers = [ "interconnect"]

If one file exceeds the maxBytes set in the config file, that file will be renamed to filename.log.1 and logs will be written to a new filename.log. This process continues for the number of files plus one (the value set in backupCount). After that point, the file being written to is rotated. The current log file is always filename.log.

For more Python configuration options, see the Python documentation at https://docs.python.org/3/library/logging.config.html.