add readiness probe logic based on node type

celestiaorg · Jun 27, 2024 · 0e8093a · 0e8093a
1 parent 739b18d
commit 0e8093a
Show file tree

Hide file tree

Showing 5 changed files with 86 additions and 7 deletions.
diff --git a/charts/celestia-node/Chart.yaml b/charts/celestia-node/Chart.yaml
@@ -21,4 +21,4 @@ maintainers:
 name: celestia-node
 sources:
   - https://github.com/celestiaorg/celestia-node
-version: 0.1.1
+version: 0.1.0
diff --git a/charts/celestia-node/templates/statefulset.yaml b/charts/celestia-node/templates/statefulset.yaml
@@ -209,18 +209,24 @@ spec:
               port: rpc
           {{- end }}
           {{- if .Values.node.customReadinessProbe }}
-          readinessProbe: {{- include "common.tplvalues.render" (dict "value" .Values.node.customReadinessProbe "context" $) | nindent 12 }}
+          readinessProbe: { { - include "common.tplvalues.render" (dict "value" .Values.node.customReadinessProbe "context" $) | nindent 12 } }
           {{- else if .Values.node.readinessProbe.enabled }}
-          readinessProbe: {{- include "common.tplvalues.render" (dict "value" (omit .Values.node.readinessProbe "enabled") "context" $) | nindent 12 }}
+          {{- if or (eq .Values.node.nodeType "bridge") (eq .Values.node.nodeType "full") }}
+          readinessProbe:
             exec:
               command:
                 - sh
                 - '-c'
                 - >
                   AUTH_TOKEN=$(celestia bridge auth read --node.store=/home/celestia 2>/dev/null);
+                  {{- if eq .Values.node.nodeType "bridge" }}
+                  curl --max-time 0.5 -s -X POST -H "Content-type: application/json" -H "Accept: application/json" -H "Authorization: Bearer $AUTH_TOKEN" -d '{"id":1,"jsonrpc":"2.0","method":"header.SyncWait","params":[]}' "http://localhost:26658";
+                  {{- else if eq .Values.node.nodeType "full" }}
                   curl --max-time 0.5 -s -X POST -H "Content-type: application/json" -H "Accept: application/json" -H "Authorization: Bearer $AUTH_TOKEN" -d '{"id":1,"jsonrpc":"2.0","method":"das.WaitCatchUp","params":[]}' "http://localhost:26658";
+                  {{- end }}
                   if [ "$?" -eq 0 ]; then exit 0; else echo "Catching up"; exit 1; fi
           {{- end }}
+          {{- end }}
           {{- if .Values.node.customStartupProbe }}
           startupProbe: {{- include "common.tplvalues.render" (dict "value" .Values.node.customStartupProbe "context" $) | nindent 12 }}
           {{- else if .Values.node.startupProbe.enabled }}

diff --git a/charts/celestia-node/values.yaml b/charts/celestia-node/values.yaml
@@ -74,6 +74,12 @@ diagnosticMode:
 ## %%MAIN_CONTAINER/POD_DESCRIPTION%%
 ##
 node:
+  ## Select from: bridge, full, light
+  ## Valid values:
+  ## - bridge
+  ## - full
+  ## - light:
+  nodeType: bridge
   ## Celestia celestia-node image
   ## ref: https://github.com/celestiaorg/celestia-node/pkgs/container/celestia-node/versions?filters%5Bversion_type%5D=tagged
   ## @param node.image.registry [default: REGISTRY_NAME] celestia-node image registry
@@ -368,7 +374,7 @@ node:
   ##
   readinessProbe:
     # -- enable readiness probe on node containers, false by default
-    enabled: false
+    enabled: true
     # -- initial delay seconds for readinessProbe, 0 by default
     initialDelaySeconds: 0
     # -- period seconds for readinessProbe, 10 by default

diff --git a/examples/celestia-node/arabica-bridge.yaml b/examples/celestia-node/arabica-bridge.yaml
@@ -0,0 +1,66 @@
+commonLabels:
+  chain_id: arabica-11
+  environment: arabica
+  network: da
+  type: bridge # check
+persistence:
+  size: 1000Gi
+  dataSource:
+    apiGroup: snapshot.storage.k8s.io
+    kind: VolumeSnapshot
+    name: da-latest
+node:
+  type: bridge # check
+  persistentVolumeClaimRetentionPolicy:
+    whenDeleted: Delete
+    whenScaled: Retain
+  otelAgent:
+    enabled: true
+    grafanaOtelSecret:
+      name: grafana-otel-token
+  resources:
+    requests:
+      cpu: 3
+      memory: 15Gi
+    limits:
+      cpu: 6
+      memory: 24Gi
+  livenessProbe:
+    enabled: true
+    periodSeconds: 10
+    failureThreshold: 3
+  readinessProbe:
+    enabled: true
+    periodSeconds: 10
+    failureThreshold: 30
+  startupProbe:
+    enabled: true
+    periodSeconds: 10
+    failureThreshold: 60
+  args:
+    - bridge  # check
+    - start
+    - --p2p.network=arabica-11
+    - --node.store=$(CELESTIA_HOME)
+    - --metrics
+    - --metrics.tls=false
+    - --p2p.metrics
+  settings:
+    address: "eyJhbGciOiJQQkVTMi1IUzI1NitBMTI4S1ciLCJjcmVhdGVkIjoiMjAyNC0wMS0wMiAxMjo1NzoyMC4yNjk1Mjg2NTQgKzAwMDAgVVRDIG09KzAuMDQwNDkyNDk0IiwiZW5jIjoiQTI1NkdDTSIsInAyYyI6ODE5MiwicDJzIjoibHBUb0pLUkFvS3NPbmRhaCJ9.SFen5RaM_7JverpZQN6VxtJ06XePE5VKjyzMyIPgqURSge-XbiLHqA.vZ4fuVFbrLODe8RM.EGKwO3uGEifnPanziF-AD4i6q_92IetSVAEz-7Grg75AJyFLV3Flt4CmfLxrJ2JHWJEKUQBNyIU81u7pd01iNzEECDy9b0jO3DYZ_5PRET8zrVPv6SKKZ_U7MLWmAe_eUajipnbqZ6NzG8UcG4qAt278ttZB1KQsKqocKlE39fSvo3iaZEzOKxS0N3tU1Xw2C3w6HgHJFk0QiP36NIHuKoP_4QJkeAomhJX-cNOYwvTd8qTFTk2-1TzLJ5DasLXsVXTUSUzCzYQhEfL8F0qFoGvuY1dWRD5BESw2Gg2oLH1w957Xd2UhRbo5ORz9POR7goEv9AZ_rWrjSV8HozUmgFwDQyM89M7oiNedEfGQoDjypzcUBcVkZLEMIdrFRpshjcpNEzy6-PZ9rSxouKb7wemG_sueX3V3gIPCTfFr6hB8_b2mV-kKeIXQRS5pJz_wXxeFHFKOHDjt3L8n5eqh.fy-zygX7rjTUqgX02sZKKg"
+    node_id: "92a499fc2ce7cabd2de89416aa0d0d07838f625d"
+    secret:
+      name: keys-da-bridge-1
+  config:
+    configtoml:
+      Core:
+        IP: consensus-validator-1
+      RPC:
+        Enabled: true
+        Address: 0.0.0.0
+      Gateway:
+        Enabled: true
+        Address: 0.0.0.0
+      Node:
+        StartupTimeout: 2m
+diagnosticMode:
+  enabled: false
diff --git a/examples/celestia-node/arabica.yaml → examples/celestia-node/arabica-full.yaml b/examples/celestia-node/arabica.yaml → examples/celestia-node/arabica-full.yaml
@@ -2,14 +2,15 @@ commonLabels:
   chain_id: arabica-11
   environment: arabica
   network: da
-  type: bridge
+  type: full # check
 persistence:
   size: 1000Gi
   dataSource:
     apiGroup: snapshot.storage.k8s.io
     kind: VolumeSnapshot
     name: da-latest
 node:
+  type: full # check
   persistentVolumeClaimRetentionPolicy:
     whenDeleted: Delete
     whenScaled: Retain
@@ -37,7 +38,7 @@ node:
     periodSeconds: 10
     failureThreshold: 60
   args:
-    - bridge
+    - full  # check
     - start
     - --p2p.network=arabica-11
     - --node.store=$(CELESTIA_HOME)
@@ -52,7 +53,7 @@ node:
   config:
     configtoml:
       Core:
-        IP: consensus-validator
+        IP: consensus-validator-1
       RPC:
         Enabled: true
         Address: 0.0.0.0