Wait for x number of scenario instances to reach a step

[mrm-gith-b]

Is there a way to synchronize a particular step so that x number of scenario instances will wait for that step to be reached before proceeding? This is a feature in some other frameworks (examples are LoadRunner and Neoload.) Using c#.

[AntyaDev]

Hi @mrm-rockport
Can you please provide a link on documentation about this feature/behaviour in LoadRunner and Neoload

[mrm-gith-b]

They are called rendezvous points for those products,
Here's an explanation for LoadRunner : https://www.perfmatrix.com/loadrunner-rendezvous-point/
and here's one for Neoload: https://documentation.tricentis.com/neoload/2024.1/en/content/reference_guide/rendezvous.htm

This is a useful feature when testing say, response time of posting an invoice with x number of simultaneous users (1,2,3 etc.)

[AntyaDev]

Got you.
Thanks.
By the way, what’s stopping you from selecting a specific endpoint and load-testing it simultaneously from multiple threads?
You could have already prepared the data to simulate that the initial dataset is ready, allowing all these threads to bombard the endpoint or functionality simultaneously.

[AntyaDev]

I suspect that this "rendezvous" negatively impacts overall latency. Any synchronization or coordination between steps means that some steps are waiting instead of bombarding, which isn’t fair when simulating a real system. I tend to believe that if you want to test a specific concurrency scenario for a particular endpoint, you should simulate it directly.

[AntyaDev]

Does it make sense?

[mrm-gith-b]

I may be misunderstanding your comments, or I've provided insufficient context.

I'm using Playwright to simulate browser access and plan on using NBomber Cluster to increase the number of browsers I can run in total.

This scenario would be several steps - login, navigate to a page, populate a form, submit the form, etc. There are Lots of backend activities on submission that take several seconds, so when several users are submitting at nearly the same time I see significant response time degradation. I know already that this is slow but I need to quantify it and then retest as we make improvements. This is why I would want to synchronize the submit step, with other uses down the road.

I've done some exploration of a solution using NATS messaging but I don't know if this is the optimal solution.

[AntyaDev]

Yeah, now I see your point about this feature—we need to think it over. I didn't know about Playwright. To be honest, you're the first client to request it! 🙂
We received a feature request to add an API for communication between agents in a cluster, so this functionality could potentially be used to implement this feature. Essentially, you need a way to signal all participants that they can now proceed with execution.

NATS will definitely be the central point for communication and signaling—that's the main reason we use it for the cluster.

Could you please submit a ticket for this? https://github.com/PragmaticFlow/NBomber/issues
Are you an NBomber Enterprise client, right?

[mrm-gith-b]

Not a client yet but I expect a license will be purchased shortly.

The attached are what I've put together (with Claude.AI), though I'm still trying to fully understand it.
NBomberExamples.zip