Add telemetry to garage staging

2 files changed, 235 insertions, 0 deletions

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
+      }
+    }
+
+  }
+}