Command Reference

mode

Specifies the running mode of the tuning program.

train, tune, recommend

--tuner-config-file, -x

Path of the core parameter configuration file of X-Tuner. The default path is xtuner.conf under the installation directory.

-

--db-config-file, -f

Path of the connection information configuration file used by the tuning program to log in to the database host. If the database connection information is configured in this file, the following database connection information can be omitted.

-

--db-name

Specifies the name of a database to be tuned.

-

--db-user

Specifies the user account used to log in to the tuned database.

-

--port

Specifies the database listening port.

-

--host

Specifies host IP address of the database instance.

-

--host-user

Specifies the username for logging in to the host where the database instance is located. The database O&M tools, such as gsql and gs_ctl, can be found in the environment variables of the username.

-

--host-ssh-port

Specifies the SSH port number of the host where the database instance is located. This parameter is optional. The default value is 22.

-

--help, -h

Returns the help information.

-

--version, -v

Returns the current tool version.

-

logfile

Path for storing generated logs.

-

output_tuning_result

(Optional) Specifies the path for saving the tuning result.

-

verbose

Whether to print details.

on, off

recorder_file

Path for storing logs that record intermediate tuning information.

-

tune_strategy

Specifies a strategy used in tune mode.

rl, gop, auto

drop_cache

Whether to perform drop cache in each iteration. Drop cache can make the benchmark score more stable. If this parameter is enabled, add the login system user to the /etc/sudoers list and grant the NOPASSWD permission to the user. (You are advised to enable the NOPASSWD permission temporarily and disable it after the tuning is complete.)

on, off

used_mem_penalty_term

Penalty coefficient of the total memory used by the database. This parameter is used to prevent performance deterioration caused by unlimited memory usage. The greater the value is, the greater the penalty is.

0 ~ 1

rl_algorithm

Specifies the RL algorithm.

ddpg

rl_model_path

Path for saving or reading the RL model, including the save directory name and file name prefix. In train mode, this path is used to save the model. In tune mode, this path is used to read the model file.

-

rl_steps

Number of training steps of the deep reinforcement learning algorithm

-

max_episode_steps

Maximum number of training steps in each episode

-

test_episode

Number of episodes when the RL algorithm is used for optimization

-

gop_algorithm

Specifies a global optimization algorithm.

bayes, pso, auto

max_iterations

Maximum number of iterations of the global optimization algorithm. (The value is not fixed. Multiple iterations may be performed based on the actual requirements.)

-

particle_nums

Number of particles when the PSO algorithm is used

-

benchmark_script

Specifies a benchmark drive script. This parameter specifies the file with the same name in the benchmark path to be loaded. Typical benchmarks, such as TPC-C and TPC-H, are supported by default.

tpcc, tpch, tpcds, sysbench ...

benchmark_path

Path for saving the benchmark script. If this parameter is not configured, the configuration in the benchmark drive script is used.

-

benchmark_cmd

Command for starting the benchmark script. If this parameter is not configured, the configuration in the benchmark drive script is used.

-

scenario

Type of the workload specified by the user

tp, ap, htap

tuning_list

List of parameters to be tuned. For details, see the share/knobs.json.template file.

-