rework for go-chi

[bigangryrobot]

there seems to be a few reasons to not use iris, here is a chi implementation

https://www.reddit.com/r/golang/comments/b481q7/a_warning_about_githubcomkatarasiris/
the-benchmarker/web-frameworks#1129
https://www.reddit.com/r/golang/comments/57w79c/why_you_really_should_stop_using_iris/

[xxx7xxxx]

Thanks for your contribution. It's totally OK to use a new web framework to handle admin API. But there's one thing I'm considering. The chi looks like a little low level by comparing with echo[1], which tries to be compatible with stdlib. How do you consider it?

[1] https://github.com/labstack/echo

[bigangryrobot]

good comments thank you both, let me clean all this up. also, ya echo would be fine and maybe thats a smaller code change overall, let me review that and see how it works out

[bigangryrobot]

Ok folks, that should address the comments you gave. Im not sure I like the logging implementation, the http helpers and lack of http tests, so I might tackle that next

[bigangryrobot]

For a performance comparison, here is a quick look using echo "GET http://localhost:2381/apis/v1/status/members"| vegeta attack -duration=25s | tee results.bin | vegeta report

Iris:

Requests      [total, rate, throughput]         1250, 50.04, 50.04
Duration      [total, attack, wait]             24.982s, 24.981s, 1.765ms
Latencies     [min, mean, 50, 90, 95, 99, max]  803.839µs, 2.136ms, 2.138ms, 2.612ms, 2.757ms, 6.506ms, 17.971ms
Bytes In      [total, mean]                     952500, 762.00
Bytes Out     [total, mean]                     0, 0.00
Success       [ratio]                           100.00%
Status Codes  [code:count]                      200:1250

Chi:

Requests      [total, rate, throughput]         1250, 50.04, 50.04
Duration      [total, attack, wait]             24.982s, 24.98s, 1.725ms
Latencies     [min, mean, 50, 90, 95, 99, max]  701.835µs, 2.154ms, 2.153ms, 2.582ms, 2.654ms, 7.62ms, 20.543ms
Bytes In      [total, mean]                     952500, 762.00
Bytes Out     [total, mean]                     0, 0.00
Success       [ratio]                           100.00%
Status Codes  [code:count]                      200:1250

And for echo "GET http://localhost:2381/apis/v1/objects"| vegeta attack -duration=25s | tee results.bin | vegeta report

Iris:

Requests      [total, rate, throughput]         1250, 50.04, 50.03
Duration      [total, attack, wait]             24.983s, 24.98s, 3.657ms
Latencies     [min, mean, 50, 90, 95, 99, max]  2.914ms, 5.539ms, 5.59ms, 6.504ms, 6.709ms, 9.21ms, 22.104ms
Bytes In      [total, mean]                     1667500, 1334.00
Bytes Out     [total, mean]                     0, 0.00
Success       [ratio]                           100.00%
Status Codes  [code:count]                      200:1250

Chi:

Requests      [total, rate, throughput]         1250, 50.04, 50.03
Duration      [total, attack, wait]             24.983s, 24.98s, 3.33ms
Latencies     [min, mean, 50, 90, 95, 99, max]  2.98ms, 5.553ms, 5.57ms, 6.426ms, 6.657ms, 9.06ms, 21.311ms
Bytes In      [total, mean]                     1667500, 1334.00
Bytes Out     [total, mean]                     0, 0.00
Success       [ratio]                           100.00%
Status Codes  [code:count]                      200:1250

[bigangryrobot]

i guess i didnt try add r.ContentLength last night 🤦 , but i added it now and it seems to work just fine so that should wrap up the logger

[benja-wu]

LGTM

[bigangryrobot]

its worth mentioning that I did roll through 80% of adding the echo framework as well, just to see how it went and it was a bit harder to implement due mainly to how the easegress framework is currently structured. It might be something to return to after some time.

[benja-wu]

its worth mentioning that I did roll through 80% of adding the echo framework as well, just to see how it went and it was a bit harder to implement due mainly to how the easegress framework is currently structured. It might be something to return to after some time.

@bigangryrobot Thank you so much for contributing this PR and providing a performance quick look. The echo is also a cool web framework but I think go-chi looks good enough for Easegress API management currently.

Feel free to open an issue about bug reportings, design discussions, or new feature proposals.

[xxx7xxxx]

Sure, I just think echo got higher abstraction than chi in some ways. But chi is fine in this simple scenario. Thanks for your contribution!