Allow logging the behaviour of middleware #248
Conversation
In a consuming application, when debugging slow responses to subscribing clients, it is useful to know more detail about how message_bus handled the request. Tracking some attributes of a connection will allow us to spot patterns in use of long polling/chunked encoding, compare response times to subscription volume, etc. Our use case involves `NewRelic::Agent.add_custom_attributes` but the implementation here is agnostic of the tool used to record this data.
|
I follow, but I kind of want to wait on us establishing the pattern in #243, it is possible a simple method extraction here would do the trick without needing to add the custom callback. |
@SamSaffron Our tests revealed that our focus particularly should be within |
|
This has been hanging so long @benlangfeld, I am going to close this off for now. |
| @@ -139,6 +139,20 @@ def handle_request(env) | |||
|
|
|||
| backlog = client.backlog | |||
|
|
|||
| if @bus.on_middleware_attributes | |||
| @bus.on_middleware_attributes.call( | |||
| messagebus_seq: client.seq, | |||
There was a problem hiding this comment.
I think this is my biggest issue with this change, would prefer if we just pass env along and a fatter object vs all these options.
In a consuming application, when debugging slow responses to subscribing clients, it is useful to know more detail about how message_bus handled the request.
Tracking some attributes of a connection will allow us to spot patterns in use of long polling/chunked encoding, compare response times to subscription volume, etc.
Our use case involves
NewRelic::Agent.add_custom_attributesbut the implementation here is agnostic of the tool used to record this data.This was a part of #243, and I think the less controversial part. I'm still working on getting something to satisfy the rest of that PR's feedback ready, but in the meantime maybe this can ship.