Configure how many engines will run when using a Taurus script (YAML)
The locations_weighted
parameter in your Taurus configuration file (yml) determines how many engines are allocated to run a test and how the Concurrency is spread over these engines.
-
Specify
locations-weighted: false
to start the given number of machines at each location and spread the load over these engines pro ratio. -
If
locations-weighted
is not specified or is set totrue
, the amount specified for each location is only used to calculate the percentage of the concurrency to run at that location.
Example 1
Running 20 users using locations-weighted: false
and 2 locations.
The result is 20 users running on 5 different engines: 3 from us-central1-a and 2 from us-east-1.
Each engine will run 4 virtual users, since we have 20 users divided into 5 engines.
concurrency: 20 locations: us-central1-a: 3 us-east-1: 2 locations-weighted: false
Example 2
Running 20 users without specifying the locations-weighted
field and 2 locations.
The result is 20 users running on 2 different engines: 3/5 of the users (12 users) will run from us-central1-a. 2/5 of the users (8 users) will run from us-east-1.
As the BlazeMeter default value allows up to 500 virtual users per engine, this example will run 1 engine per location, and a total of 2 engines for the test.
concurrency: 20 locations: us-central1-a: 3 us-east-1: 2
Example 3
Running 20 users without specifying the locations-weighted
field and 1 location.
The result is 20 users running on 1 location.
As the BlazeMeter default value allows up to 500 virtual users per engine, this example will run 1 engine per location, and a total of 1 engines for the test.
concurrency: 20 locations: us-central1-a: 3
Example 4
Running 20 users using locations-weighted: false
and 1 location.
The result is 20 users running on 2 different engines, both of them from us-central1-a.
Each engine will run 10 virtual users, since we have 20 users divided into 2 engines.
concurrency: 20 locations: us-central1-a: 2 locations-weighted: false