Skip to content
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

Get http://localhost:8080: readLoopPeekFailLocked: read tcp [::1]:60834->[::1]:8080: read: connection reset by peer #32

Closed
ghost opened this issue Jun 20, 2018 · 14 comments

Comments

@ghost
Copy link

ghost commented Jun 20, 2018

Running the example test here and 10000 users with a hatch rate of 50. I get:

readLoopPeekFailLocked: read tcp [::1]:60834->[::1]:8080: read: connection reset by peer

I have had it running at 30,000 users with one Boomer.

mac OS Sierra, 8GB, 2 cores

Thanks

@ghost
Copy link
Author

ghost commented Jun 20, 2018

Ran:

 go get github.com/myzhan/boomer

Now:

Get http://localhost:8080: net/http: HTTP/1.x transport connection broken: http: putIdleConn: too many idle connections

@myzhan
Copy link
Owner

myzhan commented Jun 21, 2018

Don't put a time.Sleep in your code, you may need boomer's max RPS support.

Try to run with --max-rps 10000.

@ghost
Copy link
Author

ghost commented Jun 21, 2018

Thanks, I noticed that last night. Any chance of a timed linear ramp-up to 10,000 RPS?

@myzhan
Copy link
Owner

myzhan commented Jun 21, 2018

@ghost
Copy link
Author

ghost commented Jun 21, 2018

Do you feel there is any value to the Locust UI now, especially users and hatch rate?

@myzhan
Copy link
Owner

myzhan commented Jun 21, 2018

You mean locustio/locust#646 ?

Locust has no support for RPS control, so far.

@ghost
Copy link
Author

ghost commented Jun 21, 2018

Sorry for the spam everyone, but I think this is what you are inferring.
https://github.com/bbc/locust-load-tests/tree/master/examples/urlbash

However, the hatch rate, Locust RPS and server RPS are out of sync.

  • 1000 users with a hatch rate of 1 in the Locust UI.
  • --max-rps 1000

@myzhan
Copy link
Owner

myzhan commented Jun 22, 2018

The hatch rate is related to the number of goroutines(users), not the RPS. In this case, 10 users or 1000 users, make no difference.

If you run with --max-rps 1000, the RPS of both locust and server, should be very close.

@ghost
Copy link
Author

ghost commented Jun 22, 2018

And they are, thank you very much. Do you think we could also increase file descriptors, decrease TCP TIME_WAIT and increase idleConnections?

@myzhan
Copy link
Owner

myzhan commented Jun 22, 2018

Only if you want to know how many long connections the server can handle, not the max requests it can process per second.

@ghost
Copy link
Author

ghost commented Jun 22, 2018

I am willing to pay a bounty to add a linear ramp-up as a command line flag (e.g.) -rps=1 which would increase to the --max-rps. adrian underscore lewis at hushmail dot com.

@myzhan
Copy link
Owner

myzhan commented Jun 23, 2018

Adding a command line flag (e.g.) request-increase-rate=1 which would increase from zero to -max-rps, or infinity if max-rps is disabled.

@ghost
Copy link
Author

ghost commented Jun 23, 2018

Hi @myzhan Shall I put the request on here: https://www.bountysource.com?

@myzhan
Copy link
Owner

myzhan commented Jun 24, 2018

Cool, thanks for letting me know such a website.

@myzhan myzhan closed this as completed Jun 24, 2018
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

1 participant