Add telemetry to garage staging

7 files changed, 330 insertions, 1 deletions

diff --git a/app/garage-staging/config/garage.toml b/app/garage-staging/config/garage.toml
index fffcf0c..ca28b79 100644
--- a/app/garage-staging/config/garage.toml
+++ b/app/garage-staging/config/garage.toml
@@ -21,3 +21,6 @@ api_bind_addr = "0.0.0.0:3990"
 bind_addr = "0.0.0.0:3992"
 root_domain = ".garage-staging-web.home.adnab.me"
 index = "index.html"
+
+[admin_api]
+bind_addr = "0.0.0.0:3909"
diff --git a/app/garage-staging/deploy/garage.hcl b/app/garage-staging/deploy/garage.hcl
index 3412a4a..44aa0c9 100644
--- a/app/garage-staging/deploy/garage.hcl
+++ b/app/garage-staging/deploy/garage.hcl
@@ -15,6 +15,7 @@ job "garage-staging" {
       port "s3" { static = 3990 }
       port "rpc" { static = 3991 }
       port "web" { static = 3992 }
+      port "admin" { static = 3909 }
     }
 
     update {
@@ -27,7 +28,7 @@ job "garage-staging" {
       driver = "docker"
 
       config {
-        image = "dxflrs/amd64_garage:v0.6.0-rc1"
+        image = "dxflrs/amd64_garage:37011a2f6ec680e4b0bbc96fa7fa86d3738d9de8"
         command = "/garage"
         args = [ "server" ]
         network_mode = "host"
@@ -120,6 +121,15 @@ job "garage-staging" {
         }
       }
 
+      service {
+        tags = [
+          "garage-staging-admin",
+        ]
+        port = 3909
+        address_mode = "driver"
+        name = "garage-staging-admin"
+      }
+
       restart { 
         interval = "30m"  
         attempts = 10  
diff --git a/app/telemetry/config/apm-config.yaml b/app/telemetry/config/apm-config.yaml
new file mode 100644
index 0000000..1c1e645
--- /dev/null
+++ b/app/telemetry/config/apm-config.yaml
@@ -0,0 +1,10 @@
+apm-server:
+  # Defines the host and port the server is listening on. Use "unix:/path/to.sock" to listen on a unix domain socket.
+  host: "0.0.0.0:8200"
+#-------------------------- Elasticsearch output --------------------------
+output.elasticsearch:
+  # Array of hosts to connect to.
+  # Scheme and port can be left out and will be set to the default (`http` and `9200`).
+  # In case you specify and additional path, the scheme is required: `http://localhost:9200/path`.
+  # IPv6 addresses should always be defined as: `https://[2001:db8::1]:9200`.
+  hosts: ["localhost:9200"]
diff --git a/app/telemetry/config/grafana/provisioning/datasources/elastic.yaml b/app/telemetry/config/grafana/provisioning/datasources/elastic.yaml
new file mode 100644
index 0000000..8108a53
--- /dev/null
+++ b/app/telemetry/config/grafana/provisioning/datasources/elastic.yaml
@@ -0,0 +1,21 @@
+apiVersion: 1
+
+datasources:
+  - name: DS_ELASTICSEARCH
+    type: elasticsearch
+    access: proxy
+    url: http://localhost:9200
+    password: ''
+    user: ''
+    database: apm-*
+    basicAuth: false
+    isDefault: true
+    jsonData:
+      esVersion: "7.10.0"
+      includeFrozen: false
+      logLevelField: ''
+      logMessageField: ''
+      maxConcurrentShardRequests: 5
+      timeField: "@timestamp"
+      timeInterval: "5s"
+    readOnly: false
diff --git a/app/telemetry/config/otel-config.yaml b/app/telemetry/config/otel-config.yaml
new file mode 100644
index 0000000..6749ee3
--- /dev/null
+++ b/app/telemetry/config/otel-config.yaml
@@ -0,0 +1,50 @@
+receivers:
+  # Data sources: metrics, traces
+  otlp:
+    protocols:
+      grpc:
+        endpoint: ":4317"
+      http:
+        endpoint: ":55681"
+  # Data sources: metrics
+  prometheus:
+    config:
+      scrape_configs:
+        - job_name: "garage"
+          scrape_interval: 5s
+          static_configs:
+            - targets:
+              - "10.42.2.21:3909"
+              - "10.42.2.22:3909"
+              - "10.42.2.23:3909"
+
+exporters:
+  logging:
+    logLevel: info
+  # see https://www.elastic.co/guide/en/apm/get-started/current/open-telemetry-elastic.html#open-telemetry-collector
+  otlp/elastic:
+    endpoint: "localhost:8200"
+    tls:
+      insecure: true
+
+processors:
+  batch:
+
+extensions:
+  health_check:
+  pprof:
+    endpoint: :1888
+  zpages:
+    endpoint: :55679
+
+service:
+  extensions: [pprof, zpages, health_check]
+  pipelines:
+    traces:
+      receivers: [otlp]
+      processors: [batch]
+      exporters: [logging, otlp/elastic]
+    metrics:
+      receivers: [otlp, prometheus]
+      processors: [batch]
+      exporters: [logging, otlp/elastic]
diff --git a/app/telemetry/deploy/docker-compose.yml b/app/telemetry/deploy/docker-compose.yml
new file mode 100644
index 0000000..576e0a9
--- /dev/null
+++ b/app/telemetry/deploy/docker-compose.yml
@@ -0,0 +1,69 @@
+version: "2"
+services:
+
+  otel:
+    image: otel/opentelemetry-collector-contrib:${OTEL_COLLECT_TAG}
+    command: [ "--config=/etc/otel-config.yaml" ]
+    volumes:
+      - ../config/otel-config.yaml:/etc/otel-config.yaml
+    network_mode: "host"
+
+  elastic:
+    image: docker.elastic.co/elasticsearch/elasticsearch:${ELASTIC_BUNDLE_TAG}
+    container_name: elastic
+    environment:
+      - "node.name=elastic"
+      - "http.port=9200"
+      - "cluster.name=es-docker-cluster"
+      - "discovery.type=single-node"
+      - "bootstrap.memory_lock=true"
+      - "ES_JAVA_OPTS=-Xms512m -Xmx512m"
+    ulimits:
+      memlock:
+        soft: -1
+        hard: -1
+      nofile: 65536
+    volumes:
+      - "es_data:/usr/share/elasticsearch/data"
+    network_mode: "host"
+
+  # kibana instance and collectors
+  # see https://www.elastic.co/guide/en/elastic-stack-get-started/current/get-started-docker.html
+  kibana:
+    image: docker.elastic.co/kibana/kibana:${ELASTIC_BUNDLE_TAG}
+    container_name: kibana
+    environment:
+      SERVER_NAME: "kibana.local"
+      # ELASTICSEARCH_URL: "http://localhost:9700"
+      ELASTICSEARCH_HOSTS: "http://localhost:9200"
+    depends_on: [ 'elastic' ]
+    network_mode: "host"
+
+  apm:
+    image: docker.elastic.co/apm/apm-server:${ELASTIC_BUNDLE_TAG}
+    container_name: apm
+    volumes:
+      - "../config/apm-config.yaml:/usr/share/apm-server/apm-server.yml:ro"
+    depends_on: [ 'elastic' ]
+    network_mode: "host"
+
+  grafana:
+    # see https://grafana.com/docs/grafana/latest/installation/docker/
+    image: "grafana/grafana:8.3.5"
+    container_name: grafana
+    # restart: unless-stopped
+    environment:
+      - "GF_INSTALL_PLUGINS=grafana-clock-panel,grafana-simple-json-datasource,grafana-piechart-panel,grafana-worldmap-panel,grafana-polystat-panel"
+    network_mode: "host"
+    volumes:
+      # chown 472:472 if needed
+      - grafana:/var/lib/grafana
+      - ../config/grafana/provisioning/:/etc/grafana/provisioning/
+
+volumes:
+  es_data:
+    driver: local
+  grafana:
+    driver: local
+  metricbeat:
+    driver: local
diff --git a/app/telemetry/deploy/telemetry.hcl b/app/telemetry/deploy/telemetry.hcl
new file mode 100644
index 0000000..bcfe3f6
--- /dev/null
+++ b/app/telemetry/deploy/telemetry.hcl
@@ -0,0 +1,166 @@
+job "telemetry" {
+  datacenters = ["neptune"]
+  type = "service"
+  
+  group "grafana" {
+    count = 1
+
+    constraint {
+      attribute = "${attr.unique.hostname}"
+      operator = "="
+      value = "cariacou"
+    }
+
+    network {
+      port "otel_grpc" {
+        static = 4317
+      }
+      port "elastic" {
+        static = 9200
+      }
+      port "kibana" {
+        static = 5601
+      }
+      port "apm" {
+        static = 8200
+      }
+      port "grafana" {
+        static = 3333
+      }
+    }
+
+    task "otel" {
+      driver = "docker"
+      config {
+        image = "otel/opentelemetry-collector-contrib:0.44.0"
+        args = [
+          "--config=/etc/otel-config.yaml",
+        ]
+        network_mode = "host"
+        ports= [ "otel_grpc" ]
+        volumes = [
+          "secrets/otel-config.yaml:/etc/otel-config.yaml"
+        ]
+      }
+
+      template {
+        data = file("../config/otel-config.yaml")
+        destination = "secrets/otel-config.yaml"
+      }
+
+      resources {
+        memory = 200
+        cpu = 100
+      }
+    }
+
+    task "elastic" {
+      driver = "docker"
+      config {
+        image = "docker.elastic.co/elasticsearch/elasticsearch:7.17.0"
+        network_mode = "host"
+        volumes = [
+          "/mnt/ssd/telemetry/es_data:/usr/share/elasticsearch/data",
+        ]
+        ports = [ "elastic" ]
+      }
+
+      resources {
+        memory = 2500
+        cpu = 500
+      }
+
+      template {
+        data = <<EOH
+node.name=elastic
+http.port=9200
+cluster.name=es-docker-cluster
+discovery.type=single-node
+bootstrap.memory_lock=true
+ES_JAVA_OPTS=-Xms512m -Xmx512m
+EOH
+        destination = "secrets/env"
+        env = true
+      }
+    }
+
+    task "kibana" {
+      driver = "docker"
+      config {
+        image = "docker.elastic.co/kibana/kibana:7.17.0"
+        network_mode = "host"
+        ports = [ "kibana" ]
+      }
+
+      template {
+        data = <<EOH
+SERVER_NAME=kibana.local
+ELASTICSEARCH_HOSTS=http://localhost:9200
+EOH
+        destination = "secrets/env"
+        env = true
+      }
+
+      resources {
+        memory = 1500
+        cpu = 500
+      }
+    }
+
+    task "apm" {
+      driver = "docker"
+      config {
+        image = "docker.elastic.co/apm/apm-server:7.17.0"
+        network_mode = "host"
+        ports = [ "apm" ]
+        args = [ "--strict.perms=false" ]
+        volumes = [
+          "secrets/apm-config.yaml:/usr/share/apm-server/apm-server.yml:ro"
+        ]
+      }
+
+      template {
+        data = file("../config/apm-config.yaml")
+        destination = "secrets/apm-config.yaml"
+      }
+
+      resources {
+        memory = 200
+        cpu = 100
+      }
+    }
+
+    task "grafana" {
+      driver = "docker"
+      config {
+        image = "grafana/grafana:8.3.5"
+        network_mode = "host"
+        ports = [ "grafana" ]
+        volumes = [
+          "/mnt/ssd/telemetry/grafana:/var/lib/grafana",
+          "secrets/elastic.yaml:/etc/grafana/provisioning/datasources/elastic.yaml"
+        ]
+      }
+
+      template {
+        data = file("../config/grafana/provisioning/datasources/elastic.yaml")
+        destination = "secrets/elastic.yaml"
+      }
+
+      template {
+        data = <<EOH
+GF_INSTALL_PLUGINS=grafana-clock-panel,grafana-simple-json-datasource,grafana-piechart-panel,grafana-worldmap-panel,grafana-polystat-panel
+GF_SERVER_HTTP_PORT=3333
+EOH
+        destination = "secrets/env"
+        env = true
+      }
+
+      resources {
+        memory = 500
+        cpu = 100
+      }
+    }
+
+  }
+}