I have service deployed through a stack on a swarm. Let's say :
someStack :
SomeServer : ...
myApplication: ....
Above this, there is a traefik server, that allows me to call various services, but also to map different urls to APIs/subservices : since the above service provide actually many sub-services such as (seen from container POV):
- /myApplication/getUsers/For/Area/51
- /myApplication/getUsers/Admins
- /myApplication/ping/enclyclopedia&code=42
- /myApplication/bricks/list&code=0937
along with other endpoints from other stacks / services (/otherApplication/toto, /yaApp/titi, etc.)
The matching endpoints being (from traefik POV) :
- /users&area=51
- /getadmins
- /ask&code=42
- /listbricks&code=0937
Theses works well... Now, Il would like to be able to perform stats on the usage of each endpoint (eg using grafana), related to myApplication overall stats . Something like :
- /users : 57 % of myApplication calls, 33 % of myApplication total response time, 15 % of myApplication total errors
- /getadmins : 33 % of myApplication call, 7% of myApplication total response time, 85 % of myApplication total errors
- /ask : 7 % of myApplication call, 40% of myApplication total response time, 0 % of myApplication total errors
- /listbricks : 3 % of cmyApplication all, 20% of myApplication total response time, 0 % of myApplication total errors
The metrics I have so far are those provided by cAdvisor and traefik itself. I am using using prometheus to pull them and build metrics on top of them. Regarding traefik's metrics, I can't see any that match my need...
I do not own 'myApplication', and thus cannot basicely implement some kind of instrumentation from the inside (or not a trivial way). I could also build metrics over traefik access logs, but I am mainly wondering if such metrics, or trick with existing metrics, could allow me to perform such stats on my application usage.
Any idea?
