[TT-15771] Prometheus pump does not emit Tyk specific metrics for OAS APIs

[ad-langebo]

Description

We encounter an issue with tyk-pump not publishing tyk specific metrics via the prometheus pump on OAS imported APIs. We use the tyk-oss helm chart deployment.

When creating a legacy TyK API via POST /tyk/apis default tyk metrics are emitted (e.g. tyk_latency_bucket, tyk_http_status_per_path, tyk_http_requests_total, etc.).

When creating an OAS API via POST /tyk/apis/oas no tyk-specific metrics are emitted whatsoever.

Sample

I have created a minimal reproducible sample here: https://github.com/ad-langebo/tyk-pump-metrics

Expected behavior

tyk-pump publishes tyk specific metrics for OAS APIs

Actual behavior

tyk-pump publishes no tyk specific metrics for OAS APIs

Versions

tyk-oss helm chart v2.2.0
tyk-gateway v5.9.2
tyk-pump v1.12.1

[andyo-tyk]

Hi @ad-langebo,

Thanks for raising this - certainly not the expected behaviour. We will take a look at this.

[andyo-tyk]

Hi @ad-langebo,

Would you please check the configuration in your Tyk OAS API definition?

From our testing, everything is working - but you need to enable traffic logs for Tyk to generate the metrics.

[ad-langebo]

Hey @andyo-tyk
according to my observations the tyk-oss helm chart automatically sets TYK_GW_ENABLEANALYTICS to true. but even after setting it explicitly, the behavior doesn't change. i still do not receive tyk metrics for the OAS api.

I have updated my repro sample with the following environment variables explicitly set for the gateway

extraEnvs: 
- name: TYK_GW_ENABLEANALYTICS
  value: "true"
- name: TYK_GW_ANALYTICSCONFIG_ENABLEDETAILEDRECORDING
  value: "true"
- name: TYK_GW_ANALYTICSCONFIG_ENABLEGEOIP
  value: "true"

edit: I also added the analytics specific OAS extensions:

"detailedActivityLogs": {
    "enabled": true
},
"detailedTracing": {
    "enabled": true
}

[ad-langebo]

Hey @andyo-tyk,
do you have another suggestion for me to try? were you able to reproduce the issue on your side?