Complete telemetry configuration

2 files changed, 42 insertions, 8 deletions

diff --git a/cluster/prod/app/telemetry/deploy/telemetry-system.hcl b/cluster/prod/app/telemetry/deploy/telemetry-system.hcl
index e4bde1a..ae9ff72 100644
--- a/cluster/prod/app/telemetry/deploy/telemetry-system.hcl
+++ b/cluster/prod/app/telemetry/deploy/telemetry-system.hcl
@@ -1,5 +1,5 @@
 job "telemetry-system" {
-	datacenters = ["neptune", "orion"]
+	datacenters = ["neptune", "orion", "bespin"]
 	type = "system"
 	priority = "100"
 
@@ -12,7 +12,7 @@ job "telemetry-system" {
 			driver = "docker"
 
 			config {
-				image = "quay.io/prometheus/node-exporter:v1.1.2"
+				image = "quay.io/prometheus/node-exporter:v1.4.0"
 				network_mode = "host"
 				volumes = [
 					"/:/host:ro,rslave"
diff --git a/cluster/prod/app/telemetry/deploy/telemetry.hcl b/cluster/prod/app/telemetry/deploy/telemetry.hcl
index d35bd7e..af3088c 100644
--- a/cluster/prod/app/telemetry/deploy/telemetry.hcl
+++ b/cluster/prod/app/telemetry/deploy/telemetry.hcl
@@ -1,9 +1,9 @@
 job "telemetry" {
-  datacenters = ["neptune"]
+  datacenters = ["neptune", "bespin"]
   type = "service"
 
   group "prometheus" {
-    count = 1
+    count = 2
 
     network {
       port "prometheus" {
@@ -11,14 +11,26 @@ job "telemetry" {
       }
     }
 
+    constraint {
+      attribute = "${attr.unique.hostname}"
+      operator = "set_contains_any"
+      value = "concombre,df-ymk"
+    }
+
     task "prometheus" {
       driver = "docker"
       config {
-        image = "prom/prometheus:v2.38.0"
+        image = "prom/prometheus:v2.39.0"
         network_mode = "host"
         ports = [ "prometheus" ]
+        args = [
+          "--config.file=/etc/prometheus/prometheus.yml",
+          "--storage.tsdb.path=/data",
+          "--storage.tsdb.retention.size=20GB",
+        ]
         volumes = [
-          "secrets:/etc/prometheus"
+          "secrets:/etc/prometheus",
+          "/mnt/ssd/prometheus:/data"
         ]
       }
 
@@ -42,6 +54,21 @@ job "telemetry" {
         destination = "secrets/consul-client.key"
       }
 
+      template {
+        data = "{{ key \"secrets/nomad/nomad.crt\" }}"
+        destination = "secrets/nomad.crt"
+      }
+
+      template {
+        data = "{{ key \"secrets/nomad/nomad-client.crt\" }}"
+        destination = "secrets/nomad-client.crt"
+      }
+
+      template {
+        data = "{{ key \"secrets/nomad/nomad-client.key\" }}"
+        destination = "secrets/nomad-client.key"
+      }
+
       resources {
         memory = 501
         cpu = 500
@@ -110,12 +137,13 @@ job "telemetry" {
     task "grafana" {
       driver = "docker"
       config {
-        image = "grafana/grafana:8.4.3"
+        image = "grafana/grafana:9.2.0"
         network_mode = "host"
         ports = [ "grafana" ]
         volumes = [
           "../alloc/data:/var/lib/grafana",
-          "secrets/prometheus.yaml:/etc/grafana/provisioning/datasources/prometheus.yaml"
+          "secrets/prometheus.yaml:/etc/grafana/provisioning/datasources/prometheus.yaml",
+          "secrets/ldap.toml:/etc/grafana/ldap.toml"
         ]
       }
 
@@ -125,9 +153,15 @@ job "telemetry" {
       }
 
       template {
+        data = file("../config/grafana-ldap.toml")
+        destination = "secrets/ldap.toml"
+      }
+
+      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=3719
+GF_AUTH_LDAP_ENABLED=true
 EOH
         destination = "secrets/env"
         env = true