Variables not passing through - FIXED

[m00nyONE]

the {{.name}} parameter in the configMap does not get passed through to prometheus. When i open the metrics tab i get no results back:

so i started investigating and wrote a very simple http-logger. When i look at the chrome console, it tells me that the correct request against argocd-metrics-server is fired: https://argo.*******.***/extensions/metrics/api/applications/http-debugger/groupkinds/pod/rows/pod/graphs/pod_cpu_line?name=http-debugger-5b54d4556c-7hp2p.*&namespace=default&application_name=http-debugger&project=administration&uid=b6e323fe-a997-431a-b81a-41f1fcbb99a0&duration=1h

so the {{.name}} is clearly there in form of "http-debugger-5b54d4556c-7hp2p.*" which is correct.

when i look at the logs of my http-logger it shows me this:

Path: /api/v1/query_range
Headers:
Host: http-debugger.default.svc.cluster.local:8080
User-Agent: Go-http-client/1.1
Content-Length: 279
Content-Type: application/x-www-form-urlencoded
Accept-Encoding: gzip
Body:
end=1706789730.0744565&query=sum(rate(container_cpu_usage_seconds_total{pod=~"http-debug-fixed-test",+image!="",+container!="POD",+container!="",+container_name!="POD"}[5m]))+by+(pod)&start=1706786130.074456&step=60
INFO:root:10.244.13.63 - - [01/Feb/2024 12:15:30] "POST /api/v1/query_range HTTP/1.1" 200 -
INFO:root:POST request,
Path: /api/v1/query_range
Headers:
Host: http-debugger.default.svc.cluster.local:8080
User-Agent: Go-http-client/1.1
Content-Length: 259
Content-Type: application/x-www-form-urlencoded
Accept-Encoding: gzip
Body:
end=1706789730.0778503&query=sum(rate(container_cpu_usage_seconds_total{pod=~"",+container!="POD",+image!="",+container!="",+container_name!="POD"}[5m]))+by+(pod)&start=1706786130.07785&step=60
INFO:root:10.244.13.63 - - [01/Feb/2024 12:15:30] "POST /api/v1/query_range HTTP/1.1" 200 -
INFO:root:POST request,

The first one is a FIXED name in the config. i created a new row just to test if its even working. And yes, it works when i hardcode the podname into the config.

The second one is with {{.name}} variable. And here we can see that it did not get passed through to prometheus ( Thanos in my case here )

So i pinned down the problem to the argocd metrics server. Then i looked into the code here but I'm not a go expert anymore :D Maybe the query parameter does not get parsed correctly? I was curious and investigated further.. i found that no placeholder whatsoever gets parsed. No namespace, no name, no whatever.

After i had a look inside the cluster to check if there is something wrong with my template, i found the solution:

Helm was the problem. I deploy all my applications via argocd and helm. So the {{.name}} got parsed by helm and replaced with "nothing"

So if anyone else has this problem, have a look here: https://github.com/helm/helm/issues/2798#issuecomment-326454957

it should look something like this:

                      {
                        "name": "pod_cpu_line",
                        "title": "CPU",
                        "description": "",
                        "graphType": "line",
                        "metricName": "pod",
                        "queryExpression": "sum(rate(container_cpu_usage_seconds_total{pod=~\"{{ "{{" }}.name{{ "}}" }}\", image!=\"\", container!=\"POD\", container!=\"\", container_name!=\"POD\"}[5m])) by (pod)"
                      },

i hope this can help someone :-)

[Tchoupinax]

Thank you!