feat: add a decoratator for monitoring method

[samsja]

Follow up of jina-ai/clip-as-service#674

Goal:

Having an decorator to monitor internal method of an Executor

class DummyExecutor(Executor):
    @requests(on='/foo')
    def foo(self, docs, **kwargs):
        self._proces(docs)
        self.proces_2(docs)

    @monitor(name='metrics_name',  documentation='metrics description')
    def _proces(self, docs):
        ...

    @monitor()
    def proces_2(self, docs):
        ...

[codecov]

Codecov Report

Merging #4737 (112e1b7) into master (f9f8600) will decrease coverage by 0.08%.
The diff coverage is 83.33%.

@@            Coverage Diff             @@
##           master    #4737      +/-   ##
==========================================
- Coverage   87.73%   87.65%   -0.09%     
==========================================
  Files         119      119              
  Lines        8661     8690      +29     
==========================================
+ Hits         7599     7617      +18     
- Misses       1062     1073      +11     

Flag	Coverage Δ
jina	87.65% <83.33%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
jina/serve/executors/__init__.py	86.36% <72.72%> (-0.39%)	⬇️
jina/serve/executors/decorators.py	97.22% <91.66%> (-1.12%)	⬇️
jina/__init__.py	65.88% <100.00%> (ø)
...deployments/config/k8slib/kubernetes_deployment.py	52.08% <0.00%> (-6.25%)	⬇️
jina/serve/runtimes/gateway/http/__init__.py	97.77% <0.00%> (-2.23%)	⬇️
...a/orchestrate/deployments/config/docker_compose.py	99.00% <0.00%> (-1.00%)	⬇️
jina/orchestrate/flow/base.py	89.57% <0.00%> (-0.61%)	⬇️
jina/helper.py	81.72% <0.00%> (-0.06%)	⬇️
jina/serve/networking.py	88.02% <0.00%> (+1.04%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f9f8600...112e1b7. Read the comment docs.

[github-actions]

Latency summary

Current PR yields:

• 🐢🐢 index QPS at 1073, delta to last 2 avg.: -18%
• 🐢🐢 query QPS at 53, delta to last 2 avg.: -30%
• 🐎🐎🐎🐎 avg flow time within 2.0012 seconds, delta to last 2 avg.: +33%
• 🐎🐎🐎🐎 import jina within 0.6085 seconds, delta to last 2 avg.: +35%

Breakdown

Version	Index QPS	Query QPS	Avg Flow Time (s)	Import Time (s)
current	1073	53	2.0012	0.6085
3.3.25	1304	79	1.6657	0.4366
3.3.24	1341	73	1.342	0.46

Backed by latency-tracking. Further commits will update this comment.

[hanxiao]

@monitor('metrics_name', 'metrics description')

1. what are the available matric names?
2. it is better to force the explicit kwargs, with *: https://stackoverflow.com/a/14298976
3. if metrics_name is basically an enum, then why the description is required, isn't the description best described by us?

[samsja]

what are the available matric names?
it is better to force the explicit kwargs, with *: https://stackoverflow.com/a/14298976
if metrics_name is basically an enum, then why the description is required, isn't the description best described by us?

'metric_name' is the name that will appear on prometheus. For example if you are monitoring the time that clip as a service spend processing some text you will call it text_preprocessing. It is not an enum for the type of the metric, therefore the description is used to give more context on what the function which is wrapped by monitoring is doing.
Btw the user can as well just call @monitor and it will produce a name and a description automatically based on the name of the function ( everything will be documented in the documentation pr)

[samsja]

@monitor('metrics_name', 'metrics description')

1. what are the available matric names?

2. it is better to force the explicit kwargs, with `*`: https://stackoverflow.com/a/14298976

3. if metrics_name is basically an enum, then why the description is required, isn't the description best described by us?

for point 2. you are right I will update it

[JoanFM]

I am missing unit tests for the monitor decorator

[samsja]

done