-
Notifications
You must be signed in to change notification settings - Fork 1.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[2/2] Make k6 cloud run --local-execution upload test archive #3931
Conversation
a6082ee
to
f99b7ad
Compare
Heads-up I tried to run the tests on my windows machine, and I can't reproduce the issue that's reported by the tests, if you folks have any ideas, I'll take it 🙇🏻 |
044d6f2
to
4d2636d
Compare
The tests are now green. There is a test in the I created a dedicated PR (#3943) to try and cater to that flakiness. Commit 4d2636d will, as a result, be dropped when merged. In the meantime, please still consider this ready for review 🦾 |
This commits aligns the handling of the --no-usage-report option in the context of the k6 cloud run command with the following agreed upon behavior: * k6 cloud run --no-usage-report should ignore the flag, and make sure we don't send the report. * k6 cloud run --local-execution --no-usage-report should take the flag in consideration, and act accordingly.
Co-authored-by: Ivan <2103732+codebien@users.noreply.github.com>
4f26cbe
to
b179a5d
Compare
|
||
exp := `{"name":"test","vus":0,"thresholds":null,"duration":0}` | ||
assert.JSONEq(t, exp, string(b)) | ||
t.Run("creating a test run should succeed", func(t *testing.T) { |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
nit: I am not a big fan of having a test with subtests jsut for grouping as you can do that with only the name of the test. Additionally it doens't add a level of identation.
In this case you also move a bunch of tests into subtests which makes the reviewing harder:(
// WithArchive is an output that can receive the archive object in order | ||
// to upload it to the cloud. | ||
type WithArchive interface { | ||
Output | ||
SetArchive(archive *lib.Archive) | ||
} | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
This is the exact thing #3282 is trying to prevent - us keep adding "abstracitons" to the Output type that only makes sense for the cloud output and don't really make it better just make it do all the work.
IMO in this case k6 cloud run --local-execution
probahly shouldn't just the output interface at all and use the cloud output directly.
Preferably it will make the test before cloud output Start
gets called so it won't actually call CreateTestRun
, but the k6 cloud run --local-execution
does.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I see your point, and I agree that it's a good idea. We actually considered a similar approach with @joanlopez. However, we encountered some challenges when trying to access the cloud output from the run entry point to handle the archiving and upload process as you described (before Start
is called). Based on my understanding of your point, I think most of our implementation matches what you describe, besides indeed how we pass down "configuration" to the cloud output.
One of the reasons for this difficulty is that the cloud run command currently relies on the k6 run command internally but forces its output to cloud by overriding the loadConfiguredTest function. This setup makes it hard to achieve what you're suggesting without the workaround we've implemented here—adding an interface that the run command can detect and use to configure the cloud output.
Do you have any suggestions on how we could accomplish this (passing down a test archive to the cloud output) without the WithArchive
interface and run
command specific logic we've introduced?
Or, are you proposing that we should define the logic for local execution directly in the cloud run command, rather than relying on the run command? This would give us more flexibility and control, potentially avoiding changes to the output interface and the run command, as discussed in #3282, but would result in a much bigger and riskier change.
Please let me know if I’ve interpreted your suggestion correctly 🙇🏻
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unfortunately I don't have a 10 line fix that will make this possible.
As explained in the linked issue, this will only get worse and for logs AFAIK practically impossible to be done without refactoring how we run k6.
At this point it seems that the code in the "run" command will need to be refactored to be more composable and then reused in k6 cloud run --local-execution
.
Looking at the cmd
package it seems like it might be a good idea if the cmd
packages deals more wiht the configuraiton and parsing of cli flags and less wiht actually running the k6 app. Currently both of those are very intertwined but that likely is not a good idea.
As mentioned privately I am okay with us merging this for the purposes of getting this released in 0.54.0, but will definitely prefer if work on the actual refactoring and working towards the mentioned issue goals of stop making the output interface just one big pile of one off editions and the cloud output in an omnipotent part of the code.
@@ -68,6 +68,13 @@ type WithThresholds interface { | |||
SetThresholds(map[string]metrics.Thresholds) | |||
} | |||
|
|||
// WithArchive is an output that can receive the archive object in order | |||
// to upload it to the cloud. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
// to upload it to the cloud. | |
// to upload it to the cloud. | |
// NOTE: This API is EXPERIMENTAL and may be changed, renamed or | |
// completely removed in a later k6 release. |
What?
This Pull request modifies the behavior of the
k6 cloud run
command when using the--local-execution
mode. It makes uploading the test archive to the cloud a default. This is meant to support our internal efforts to improve the UX of our cloud offering.This Pull Request also adds support for an additional
--no-archive-upload
flag to thek6 cloud run
command, which enables users to prohibit uploading their test archive to our cloud.Review
Archive upload
Following a recently published change in our platform, creating test runs now supports providing the test archive along the test run parameters. To provide an archive when creating a test run, the request is expected to be a multipart form instead of the traditional JSON body. Creating test runs without archives does not change and works as before. To support this change you will find that we had to modify some of the behavior of the cloud API client
CreateTestRun
method.Configuration aspects
We developed this PR hand-in-hand with @joanlopez and particularly struggled to set up its configuration aspects. We took the approach that felt the most convenient and practical to us, in lack of a better approach. If you have ideas or insights on how to handle that particular part specifically, please let us know 👍🏻
Checklist
make lint
) and all checks pass.make tests
) and all tests pass.Related PR(s)/Issue(s)
#3904