Replace httpx with request for sync version

[sacOO7]

• Related implementation https://github.com/ably/ably-python/pull/200/files
• This is due to the fact that, httpx causes performance issues when hitting limit of 100k requests/minute.
• Link to internal discussion > https://ablyreal-time.slack.com/archives/C030C5YLY/p1698238346488369?thread_ts=1697535419.670029&cid=C030C5YLY

┆Issue is synchronized with this Jira Task by Unito

[sync-by-unito[bot]]

➤ Automation for Jira commented:

The link to the corresponding Jira issue is https://ably.atlassian.net/browse/SDK-3914

[sacOO7]

TODO

• [ ] 1. Reproduce the issue using load testing on a dedicated server.
• [ ] 2. If reproduced, then make a change to switch to a different library.

@owenpearson has an idea about how to reproduce the issue and do the load testing.

[sacOO7]

Some more findings ->

• Current SDK uses httpx version as 0.24.1 which has dependency on httpcore uses -> 0.17.3
• Latest version of httpx is 0.25.2, it uses httpcore -> 1.0.2
• Some critical bugfixes has been addressed in latest version of httpx.
• While running tests make sure we run it on both older and newer versions of httpx.

[sacOO7]

One of the related fix seems to be -> https://github.com/encode/httpcore/pull/823

[sacOO7]

The current SDK only supports httpx to be used in http2 mode -> https://github.com/ably/ably-python/blob/10cb4a08c64f915a2878ffd1139b08a50811124d/ably/http/http.py#L132. There is still an active issue for http2 -> https://github.com/encode/httpcore/issues/775 that talks about gracefully closing the connection. Fix is already done for http1.1.

[sacOO7]

NEW TODO

• [ ] 1. Reproduce the issue using load testing on a dedicated server.
  • [ ] httpx -> 0.24.1 and http2 ( use current SDK as is ! )
  • [ ] httpx -> 0.24.1 and http1.1
  • [ ] httpx -> 0.25.2 and http2
  • [ ] httpx -> 0.25.2 and http1.1
• [ ] 2. If still facing the issue, then make a change to requests library that uses http 1.1 rather than http2 😒

[Ousret]

I don't know whether it could help your case, but you may consider taking a look at Niquests instead of downgrading to requests. Your feedback would be interesting. Would be happy to get this project where you aim to reach.

[sacOO7]

@Ousret Thanks for the suggestion !

[sacOO7]

Seems we can use https://locust.io/ as a tool for load testing.

[sacOO7]

https://medium.com/@yusufenes3494/how-to-build-your-own-load-test-a-step-by-step-guide-1a8367f7f6a2

Seems locusts by default uses python-requests for executing test against given server https://docs.locust.io/en/stable/increase-performance.html. No wonder, python-requests is stable and doesn't cause issue on high throughput.

Need to check if we can configure other client like httpx instead of default python-requests

[sacOO7]

One of the way is to clone python-locust project and modify https://github.com/locustio/locust/blob/master/locust/clients.py file to use httpx client instead of python-requests. This will help us to use all features of locust client like result tracking, without writing explicit code for the same.