Requests Per Second Plot Breaks When There are too Many Unique URLs

[williamlhunter]

Description of issue

Plotting requests per second breaks when locust has visited around 500 unique URLs

Expected behavior

Total requests per second works as long as needed

Actual behavior

It seems to break once you hit too many unique URLs. I believe this is because locust gets the stats for every URL every time it updates the web ui.

Environment settings

• OS: Windows
• Python version: 3.7.3
• Locust version: 0.10.0

Steps to reproduce (for bug reports)

I am using locust to test a retail point of sale rest api. The api uses document SIDs in the url, so everytime I make a new transaction, I get 3-4 new urls. Not only does this make analyzing the results harder, as I need to group the urls together with regular expressions in pandas, but i believe that it is breaking the plotting of requests per second as well.
I can hammer the api with requests that don't use the document SID in the url forever, but I can only run a few hundred transactions before the plotting breaks.
To make sure, I made 15 new receipts per second for about 30 mins and everything worked correctly. Afterword, I changed my locust file to make a blank receipt and then request it's contents (generating a request on a new url) and was only able to run the load test for 2 mins before plotting broke. (at around 500 unique urls)

Possible fix

I would love to be able to set the field that requests are grouped by to a string. Something like:
response = self.client.post("/v1/rest/document" headers=self.header, json=payload, tag="creating a document")
but I'm not incredibly experienced, so this may not be as easy to implement as I'm hoping.
Also this implies that I'm correct as to why the requests per second plot is breaking. Furthermore, this would save me (and people doing similar testing) from having to group api calls after running a test.

[cgoldberg]

before the plotting breaks

what about it is broken?

[williamlhunter]

Sorry, I was in a bit of a rush and totally forgot to add that. Here's an example:

As you can see, the plot shows no data for each point in the second half. However, the rps count at the top is perfectly functional.

[cgoldberg]

please give a clear description of the issue, including details of what you are expecting and what is actually happening. Please exclude all other information, conjectures, or ideas for possible fixes. This issue report is packed with information that is completely irrelevant... which makes it difficult to follow.

(For example... I can't even tell if this is a bug report, or a feature request for tagging responses ... it seems like both)

[williamlhunter]

Fair, I'm a bit new to this if you can't tell.

Primarily, I want to fix the issue, but I believe that tagging would conveniently solve the bug as well. Here is a revised bug report:

Description

Requests per second stops plotting under certain circumstances.

Expected behavior

The reported RPS in the locust web client should be graphed accurately in the total requests per second chart.

Actual behavior

Total Requests per second is not graphing properly under certain circumstances.

Generating Unique URLs:

Chart

JSON

Repeatedly hitting the same URL:

Chart

JSON

Environment Settings

• OS: Windows and Arch Linux (running in a vm)
• Python version: 3.7.3
• Locust Version: 0.11.0

Steps to reproduce

Write a locust file that makes requests on a large number of unique URLs. Once the number of requested URLs hits about 500, the graph breaks.

[williamlhunter]

At risk of making more conjecture, here is where I believe the bug originates:

I believe the red bit is being truncated. This is in web.py, starting at line 104

[RyanW89]

Have you tried adding name="string" to your request? ie
response = self.client.post("/v1/rest/document" headers=self.header, json=payload, name="creating a document")

unless I am mistaken, this is what you are after.

[williamlhunter]

Wow, thanks for pointing that out. That's exactly the feature I was looking for.

However, I'm going to leave this issue open because the total requests per second plot still breaks above 500 urls.

[heyman]

Fixed by #1060 if I'm not mistaken.