Testing threads
A load generator creates parallel testing threads. Operations within threads are performed sequentially, that is, after a request is created, a response is awaited. The thread where a response is awaited is unavailable for new requests. To ensure the required load, a generator uses the following threads.
You can track thread utilization on a chart in the report.
The maximum number of requests per second depends on the number of testing threads and the total request response time in a given second:
RPS_max = number of testing threads / (total request response time in a given second)
Where RPS_max is the maximum number of requests per second.
By default, Phantom creates 1000 testing threads.
Pandora creates 1000 testing threads unless you filled in the Testing threads field when configuring via the form. However, if the configuration file was used and the number of testing threads was not specified, Pandora will terminate with an error.
If you need to set the number of threads, configure it in the management console.
You can calculate the required number of testing threads using this formula:
Number of testing threads = RPS x (average response time in seconds)
Where RPS is the number of requests per second.
For example, to ensure a load of 200 requests per second at an average response time of 50 ms, you need to create 10 testing threads.
Sample configuration file
An example of configuring testing threads to generate 200 requests per second over 5 minutes:
pandora:
enabled: true
package: yandextank.plugins.Pandora
config_content:
pools:
- id: HTTP
gun:
type: http
target: localhost:80
ammo:
file: /tmp/ammofile
type: uri
result:
type: phout
destination: ./phout.log
startup:
type: once
times: 10 # Number of threads
rps:
- duration: 300s # test duration
type: const # load type
ops: 200 # number of requests per second
log:
level: error
monitoring:
expvar:
enabled: true
port: 1234
phantom:
enabled: true
package: yandextank.plugins.Phantom
address: localhost:80
ammo_type: uri
instances: 10 # Number of threads
load_profile:
load_type: rps
schedule: const(200,5m) # load schedule: 200 requests per second over 5 minutes
uris: ['/']