Skip to content

Latest commit

 

History

History
115 lines (93 loc) · 3.99 KB

tsbs_load.md

File metadata and controls

115 lines (93 loc) · 3.99 KB
Raw

Supplemental Guide for tsbs_load

The tsbs_load executable can benchmark data ingestion for all the implemented databases.

Generating a config file

tsbs_load uses YAML files to specify the configuration for running the load benchmark.

The config file is separated in two top-level sections:

data-source:
  ...
loader: 
  ...
  • data-source contains the configuration for where to read the data from (type: SIMULATOR or type: FILE)
    • For SIMULATOR the configuration specifies the time range to be simulated, the use-case, scale and other properties that regard the data
    • For FILE the configuration only specifies the location of the pre-generated file with tsbs_generate_data
  • loader contains the configuration for the loading the data. Two sub-sections are important here db-specific and runner
    • The db-specific configuration varies depending of the target database and for TimescaleDB contains information about user, password, ssl mode, while for influx it contains information about backoff interval, replication factor etc.
    • The runner configuration specifies the number of concurrent workers to use, batch size, hashing and so on

To generate an example configuration file for a specific database run

$ tsbs_load config --target=<db-name> --data-source=[FILE|SIMULATOR]

specifying db-name to one of the implemented databases and data-source to FILE or SIMULATOR

⚠️ The generated config file will be populated with the default values for each property.

The generated config file is saved in ./config.yaml

Sample config files

You can find sample YAML configuration files for TimescaleDB in the sample-configs directory. Both single and multi-node examples are provided for FILE and SIMULATOR modes.

On the fly simulation and load with data-source: SIMULATOR

When you run tsbs_generate_data a simulator is created for the selected use case and the simulated data points are serialized to a file. tsbs_load utilizes the same simulators but the simulated points are directly piped to the worker clients that send batches of data to the databases.

You can notice that the same properties you configure in the YAML file are the same flags that you need to specify when running tsbs_generate_data.

You can run tsbs_load with

$ tsbs_load load <db_name> --config=./path-to-config.yaml

Where <db_name> is one of the implemented databases or you can run

$ tsbs_load load --help

for a list of the available databases.

Information about a property and overriding

The generated yaml file with tsbs_load config does not contain information about what each of the properties represents. You can easily discover more details about each property by running:

$ tsbs_load load --help

This will list all the available flags configurable for all databases. These flags include the flags for data-source and loader.runner. The --loader.runner.db-name flag corresponds to the property:

loader:
  runner:
    db-name: some-db

in the YAML config file. With the type, description, and default value next to the flag name as :

string, Name of database (default "benchmark")

Information about database specific flags

Some of the properties are only valid for specific databases. These properties go under the loader.db-specific section. To view information about them you can run:

$ tsbs_load load <db_name> --help

For example for timescaledb, you can see the following:

$ tsbs_load load timescaledb --help
...
--loader.db-specific.chunk-time 
    duration
    Duration that each chunk should represent, e.g., 12h (default 12h0m0s)
...

Overriding values

  • Each property has a default value, used if not otherwise overridden
  • An entry in the config YAML file overrides the default value
  • A flag passed at runtime overrides an entry in the YAML file