Closed Sirozha1337 closed 4 years ago
As far as I know it's the desired behaviour.
Increased duration of the test with an open load (load_type: rps
) usually means that at some point tank/phantom hit the limit of instances (defaults to 1000) and could not produce planned amount of requests per second. Additional test time is spent to send that "unsent" requests.
1) In order to solve the root cause of your problem you can increase the number of instances (instances: N
) using the approximate Littles Law formula:
<Number of instances> = <Max RPS> * <Max response time, sec>
and use multiple generators if number of instances is too large (20+k).
2) In order to solve the symptoms (longer test duration) you can configure autostop criteria as follows:
autostop:
enabled: true
autostop:
- limit(<desired test duration (e.g. 10m)>)
@afilatov, second solution sounds just like what I was thinking about. Seems like I somehow missed that part of the documentation or maybe it wasn't there a year ago. Anyways, I think this issue can be closed now. Thank you for your feedback!
schedule: const(100000, 10m)
It seems like a bug to me, but feel free to close it, if that's the desired behaviour.