[DISCUSS] Should we add metric for log sender and zipkin ?

[gy09535]

Issue description

Now the log sender batch store the data in memory , we can not know how many logs or spans is send, so I think we can add some metric to monitor it.

Environment

• apisix version (cmd: apisix version):
• OS:

[gy09535]

I do it in my private repository should apisix need it?

[spacewander]

Could you provide a design about it?

[spacewander]

I think we need to allow plugin to register their metrics. Like a _M.collect method and prometheus will call them when in need.

[gy09535]

Could you provide a design about it?

When prometheus exporter init , I add a gauge to metric with label : node，route_id, type。And I export two module function inc and set to change the gauge value. In "syslog-plugin" log method, I use this module method inc it value, and in batch process I reduce it value. And zipkin is also do that.
This can help us know how many remaining spans and logs per route can not send.

[gy09535]

I think we need to allow plugin to register their metrics. Like a _M.collect method and prometheus will call them when in need.

We can add plugin init method to register their own metric,And provide api to change it value.

[spacewander]

We can add plugin init method to register their own metric,And provide api to change it value.

@gy09535
Pull request is welcome!

[gy09535]

We can add plugin init method to register their own metric,And provide api to change it value.

@gy09535
Pull request is welcome!

haha, I wil do it.

[gy09535]

We can add plugin init method to register their own metric,And provide api to change it value.

@gy09535
Pull request is welcome!

Please assign to me.