Small cleanup

example/manage.nomad.hcl

@@ -9,8 +9,7 @@ job "postgres-manage" {
 
   datacenters = ["dc1"]
   region      = "global"
-
+  priority    = 80
-
 
 
   group "manage" {

example/postgres.nomad.hcl

@@ -6,38 +6,40 @@ job "postgres" {
   priority    = 80
 
 
-  constraint {
-    operator = "distinct_hosts"
-    value    = "true"
-  }
-
-
-
   group "server" {
 
+    count          = 1
+    shutdown_delay = "6s"
+
+
+    constraint {
+      operator = "distinct_hosts"
+      value    = "true"
+    }
+
+
+    update {
+      healthy_deadline  = "2h"
+      progress_deadline = "3h"
+    }
+
     ephemeral_disk {
       # Use minimal ephemeral disk
       size = 101
     }
 
-    count = "1"
+
     network {
       mode = "bridge"
-      # Patroni API for node to check each others
+      # Patroni API for nodes to check each others
       port "patroni" {
         to = 8080
       }
-      # When running with patroni, nodes must reach each others postgres service
+      # When running with patroni, nodes must reach each others postgres service, so we expose a port
       port "postgres" {
         to = 5432
       }
     }
-
-    update {
-      # Set super high deadlines as recovery can take lots of time
-      healthy_deadline  = "48h"
-      progress_deadline = "72h"
-    }
     service {
       name = "postgres"
       port = 5432
@@ -121,12 +123,14 @@ job "postgres" {
     }
 
 
+
     volume "backup" {
       source          = "postgres-backup"
       type            = "csi"
       access_mode     = "multi-node-multi-writer"
       attachment_mode = "file-system"
     }
+
     volume "data" {
       source          = "postgres-data"
       type            = "csi"

postgres.nomad.hcl

@@ -1,40 +1,32 @@
 job "[[ .instance ]]" {
 
-[[- $c := merge .pg.server .pg . ]]
+[[- $c := merge .pg . ]]
 
 [[ template "common/job_start" $c ]]
 
   group "server" {
+[[- $c := merge $c.server $c ]]
 
-    ephemeral_disk {
+[[- /* Ensure count is 1 when in recovery mode */]]
-      # Use minimal ephemeral disk
+[[ template "common/group_start" merge (dict "count" (ternary 1 $c.count $c.recovery)) $c ]]
-      size = 101
-    }
 
-    count = "[[ $c.recovery | ternary 1 $c.count ]]"
     network {
       mode = "bridge"
 [[- if conv.ToBool $c.prometheus.enabled ]]
       port "metrics" {}
       port "patroni-metrics" {}
 [[- end ]]
-      # Patroni API for node to check each others
+      # Patroni API for nodes to check each others
       port "patroni" {
         to = 8080
       }
-      # When running with patroni, nodes must reach each others postgres service
+      # When running with patroni, nodes must reach each others postgres service, so we expose a port
       port "postgres" {
         to = 5432
       }
     }
 
-    update {
+[[- if not $c.recovery ]]
-      # Set super high deadlines as recovery can take lots of time
-      healthy_deadline  = "48h"
-      progress_deadline = "72h"
-    }
-
-[[- if not .pg.server.recovery ]]
     service {
       name = "[[ .instance ]][[ $c.consul.suffix ]]"
       port = 5432
@@ -116,9 +108,9 @@ job "[[ .instance ]]" {
       config {
         image      = "[[ $c.image ]]"
         # Set shm_size to half of the total size
-        shm_size   = [[ mul .pg.server.resources.memory 524288 ]]
+        shm_size   = [[ mul $c.resources.memory 524288 ]]
         volumes    = ["local/mkdir-socket.sh:/entrypoint.d/70-mkdir-socket.sh"]
-[[- if .pg.server.recovery ]]
+[[- if $c.recovery ]]
         # Running in recovery mode : just start the container and wait
         # so we can enter it and manually recover what we need to
         command    = "sleep"

variables.yml

@@ -6,6 +6,10 @@ instance: postgres
 
 pg:
 
+  nomad:
+    # Set higher priority for the postgres job
+    priority: 80
+
   vault:
     pki:
       ou: Postgres
@@ -61,12 +65,14 @@ pg:
       PGBACKREST_STANZA: '[[ .instance ]]'
 
     nomad:
-      # Set higher priority for the postgres job
-      priority: 80
       # Enforce running on distinct hosts
       constraints:
         - operator: distinct_hosts
           value: true
+      update:
+        # When running in recovery mode, use huge deadlines as it can take a lot of time
+        healthy_deadline: '[[ .pg.server.recovery | ternary "48h" "2h" ]]'
+        progress_deadline: '[[ .pg.server.recovery | ternary "72h" "3h" ]]'
 
     # In recovery mode, neither patroni nor postgres will be started. The container will start and wait for manual recovery
     recovery: false