supports to logging apisix_latency and upstream_latency by modifying the full log format

[jagerzhang]

I hope apisix can logging apisix_latency and upstream_latency to help us with request analysis: #5910

At present, it is known that it can be implemented by modifying the full log format with existing logic code (by kyroslin):

jagerzhang@9c96dba

This should be a very general requirement, Please have a look.

[tzssangglass]

cc @spacewander

From my point of view, this is great. But so far I've only seen you bring it up. Let's wait a little time, how about if more people call for this need and let's activate it?

[jagerzhang]

ok

[jamesislebron]

can't agree more!
upstream_latency will help a lot for tracing error.

[spacewander]

@jagerzhang @jamesislebron
Could you provide a detailed use case that why this feature is necessary to you?

[jamesislebron]

My api users/customers always report that there are some requests wasted a long time;

With the apisix_latency and upstream_latency log, I can find out the most time happened on apisix or upstream conveniently.

[jagerzhang]

I heard from the community that access log for apisix also has these two fields, so I think the logs reported by plug-in should also have these fields.

in addition, i think this feature:

• does not affect performance
• references off the exists code
• enriches the log, meet the needs of operation analysis, especially when the user feedback request takes too long

In short, small changes bring big benefits.

[spacewander]

I see. So there are strong reasons to add it to the default format.
PR is welcome!

[jagerzhang]

#6063