From 1ea704f3304cda823780e4a6cab504b854eb035c Mon Sep 17 00:00:00 2001
From: Lianne Li <lmlianne@outlook.com>
Date: Mon, 26 Aug 2024 14:12:45 +0800
Subject: [PATCH] [CELEBORN-1552] automatically support prometheus to scrape
 metrics for helm chart

### What changes were proposed in this pull request?
1. Add Annotations to Master Service and Worker Service for automatically scraping by Prometheus.
2. Add Ports to Worker Service, since it's empty before that prometheus cannot connect to workers.

### Why are the changes needed?
Although master and worker provide http interfaces, We still need add annotations manually for prometheus automatically scraping.

### Does this PR introduce _any_ user-facing change?
No. Users will not feel any changes to install and use Celeborn.

### How was this patch tested?
test locally and in dev environment.

Before:
![image](https://github.com/user-attachments/assets/d924929f-1cd9-4487-afc6-08390fc8dfc2)

After:
![image](https://github.com/user-attachments/assets/145b0727-e66a-4268-af4d-cf0619eb3b14)

Closes #2673 from lianneli/PR-1552.

Authored-by: Lianne Li <lmlianne@outlook.com>
Signed-off-by: Shuang <lvshuang.xjs@alibaba-inc.com>
---
 charts/celeborn/templates/_helpers.tpl        | 75 +++++++++++++++++++
 charts/celeborn/templates/master/service.yaml |  2 +
 charts/celeborn/templates/worker/service.yaml |  3 +
 3 files changed, 80 insertions(+)

diff --git a/charts/celeborn/templates/_helpers.tpl b/charts/celeborn/templates/_helpers.tpl
index 6224a816c..770112936 100644
--- a/charts/celeborn/templates/_helpers.tpl
+++ b/charts/celeborn/templates/_helpers.tpl
@@ -155,3 +155,78 @@ Create the name of the worker podmonitor to use
 {{- define "celeborn.workerPodMonitorName" -}}
 {{ include "celeborn.fullname" . }}-worker-podmonitor
 {{- end }}
+
+{{/*
+Create master annotations if metrics enables
+*/}}
+{{- define "celeborn.masterMetricsAnnotation" -}}
+{{- $metricsEnabled := true -}}
+{{- $metricsPath := "/metrics/prometheus" -}}
+{{- $masterPort := 9098 -}}
+{{- range $key, $val := .Values.celeborn }}
+{{- if eq $key "celeborn.metrics.enabled" }}
+{{- $metricsEnabled = $val -}}
+{{- end }}
+{{- if eq $key "celeborn.metrics.prometheus.path" }}
+{{- $metricsPath = $val -}}
+{{- end }}
+{{- if eq $key "celeborn.master.http.port" }}
+{{- $masterPort = $val -}}
+{{- end }}
+{{- end }}
+{{- if eq (toString $metricsEnabled) "true" -}}
+prometheus.io/path: {{ $metricsPath }}
+prometheus.io/port: '{{ $masterPort }}'
+prometheus.io/scheme: 'http'
+prometheus.io/scrape: 'true'
+{{- end }}
+{{- end }}
+
+{{/*
+Create worker annotations if metrics enables
+*/}}
+{{- define "celeborn.workerMetricsAnnotation" -}}
+{{- $metricsEnabled := true -}}
+{{- $metricsPath := "/metrics/prometheus" -}}
+{{- $workerPort := 9096 -}}
+{{- range $key, $val := .Values.celeborn }}
+{{- if eq $key "celeborn.metrics.enabled" }}
+{{- $metricsEnabled = $val -}}
+{{- end }}
+{{- if eq $key "celeborn.metrics.prometheus.path" }}
+{{- $metricsPath = $val -}}
+{{- end }}
+{{- if eq $key "celeborn.worker.http.port" }}
+{{- $workerPort = $val -}}
+{{- end }}
+{{- end }}
+{{- if eq (toString $metricsEnabled) "true" -}}
+prometheus.io/path: {{ $metricsPath }}
+prometheus.io/port: '{{ $workerPort }}'
+prometheus.io/scheme: 'http'
+prometheus.io/scrape: 'true'
+{{- end }}
+{{- end }}
+
+{{/*
+Create worker Service http port params if metrics enables
+*/}}
+{{- define "celeborn.workerServicePort" -}}
+{{- $metricsEnabled := true -}}
+{{- $workerPort := 9096 -}}
+{{- range $key, $val := .Values.celeborn }}
+{{- if eq $key "celeborn.metrics.enabled" }}
+{{- $metricsEnabled = $val -}}
+{{- end }}
+{{- if eq $key "celeborn.worker.http.port" }}
+{{- $workerPort = $val -}}
+{{- end }}
+{{- end }}
+{{- if eq (toString $metricsEnabled) "true" -}}
+ports:
+  - port: {{ $workerPort }}
+    targetPort: {{ $workerPort }}
+    protocol: TCP
+    name: celeborn-worker-http
+{{- end }}
+{{- end }}
\ No newline at end of file
diff --git a/charts/celeborn/templates/master/service.yaml b/charts/celeborn/templates/master/service.yaml
index aee0ce1df..73a7c229b 100644
--- a/charts/celeborn/templates/master/service.yaml
+++ b/charts/celeborn/templates/master/service.yaml
@@ -21,6 +21,8 @@ metadata:
   name: {{ include "celeborn.masterServiceName" . }}
   labels:
     {{- include "celeborn.labels" . | nindent 4 }}
+  annotations:
+    {{- include "celeborn.masterMetricsAnnotation" . | nindent 4 }}
 spec:
   selector:
     {{- include "celeborn.selectorLabels" . | nindent 4 }}
diff --git a/charts/celeborn/templates/worker/service.yaml b/charts/celeborn/templates/worker/service.yaml
index 2f8b1ec24..f56483054 100644
--- a/charts/celeborn/templates/worker/service.yaml
+++ b/charts/celeborn/templates/worker/service.yaml
@@ -21,9 +21,12 @@ metadata:
   name: {{ include "celeborn.workerServiceName" . }}
   labels:
     {{- include "celeborn.labels" . | nindent 4 }}
+  annotations:
+    {{- include "celeborn.workerMetricsAnnotation" . | nindent 4 }}
 spec:
   selector:
     {{- include "celeborn.selectorLabels" . | nindent 4 }}
     app.kubernetes.io/role: worker
   type: {{ .Values.service.type }}
   clusterIP: None
+  {{- include "celeborn.workerServicePort" . | nindent 2 }}