XS⚠️ ◾ fix: liveness probe tcpSocket default to prevent Raft catch-up restart loops

Chart.yaml

@@ -1,9 +1,11 @@
 apiVersion: v2
 name: typesense
 description: >-
-  Deploy Typesense search engine on Kubernetes with Raft-based HA clustering, Prometheus metrics, and Gateway API support. This chart is not officially maintained by or affiliated with the Typesense project.
+  Deploy Typesense search engine on Kubernetes with Raft-based HA clustering,
+  Prometheus metrics, and Gateway API support. This chart is not officially
+  maintained by or affiliated with the Typesense project.
 type: application
-version: 1.1.0
+version: 1.1.1
 appVersion: "30.1"
 icon: https://typesense.org/typesense-logo.svg
 home: https://github.com/hackthebox/typesense-helm
@@ -36,5 +38,9 @@ annotations:
     - name: metrics-exporter
       image: imatefx/typesense-prometheus-exporter:v0.1.5
   artifacthub.io/changes: |
-    - kind: added
-      description: Initial open-source release
+    - kind: fixed
+      description: "liveness probe default changed to tcpSocket to prevent restart loops during Raft catch-up"
+    - kind: changed
+      description: "probe type is now controlled via .type field (httpGet or tcpSocket) to avoid Helm merge producing multiple handlers"
+    - kind: changed
+      description: "default CPU request set to 2000m per Typesense minimum requirements"

README.md

@@ -1,6 +1,6 @@
 # typesense
 
-![Version: 1.1.0](https://img.shields.io/badge/Version-1.1.0-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: 30.1](https://img.shields.io/badge/AppVersion-30.1-informational?style=flat-square)
+![Version: 1.1.1](https://img.shields.io/badge/Version-1.1.1-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: 30.1](https://img.shields.io/badge/AppVersion-30.1-informational?style=flat-square)
 
 [![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/typesense)](https://artifacthub.io/packages/search?repo=typesense)
 
@@ -233,9 +233,13 @@ storage:
 | ingress.enabled | bool | `false` | Enable or disable Ingress for the application |
 | ingress.hosts | list | `[]` | List of hostnames the Ingress will route traffic for |
 | ingress.prefix | string | `"/"` | The URL path prefix for the application |
-| livenessProbe.failureThreshold | int | `2` | Number of failed liveness checks before restarting the container |
-| livenessProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense health for liveness |
-| livenessProbe.periodSeconds | int | `10` | Period (in seconds) to perform the liveness check |
+| livenessProbe.failureThreshold | int | `6` | Number of failed liveness checks before restarting the container |
+| livenessProbe.httpGet.path | string | `"/health"` |  |
+| livenessProbe.httpGet.port | string | `"http"` |  |
+| livenessProbe.periodSeconds | int | `20` | Period (in seconds) to perform the liveness check |
+| livenessProbe.tcpSocket.port | string | `"http"` |  |
+| livenessProbe.timeoutSeconds | int | `3` |  |
+| livenessProbe.type | string | `"tcpSocket"` | Probe type: 'tcpSocket' or 'httpGet'. Default is tcpSocket to check only that the process is alive and listening. Avoid httpGet /health for liveness: Typesense returns 503 when the write queue exceeds --healthy-write-lag (default 500), which occurs normally during Raft catch-up after a restart, causing a liveness-triggered restart loop that prevents the cluster from ever recovering. |
 | metrics.enabled | bool | `false` | Enable Prometheus metrics sidecar |
 | metrics.image.pullPolicy | string | `"IfNotPresent"` | Image pull policy for metrics exporter |
 | metrics.image.repository | string | `"imatefx/typesense-prometheus-exporter"` | Metrics exporter image repository |
@@ -255,11 +259,15 @@ storage:
 | podSecurityContext.runAsGroup | int | `3000` | Group ID for running the Typesense process |
 | podSecurityContext.runAsNonRoot | bool | `true` | Ensure the container does not run as root |
 | podSecurityContext.runAsUser | int | `10000` | User ID for running the Typesense process |
-| readinessProbe.failureThreshold | int | `3` | Number of failed readiness checks before marking the pod as unready |
-| readinessProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense readiness |
-| readinessProbe.periodSeconds | int | `5` | Period (in seconds) to perform the readiness check |
+| readinessProbe.failureThreshold | int | `12` | Number of failed readiness checks before marking the pod as unready |
+| readinessProbe.httpGet.path | string | `"/health"` |  |
+| readinessProbe.httpGet.port | string | `"http"` |  |
+| readinessProbe.periodSeconds | int | `10` | Period (in seconds) to perform the readiness check |
+| readinessProbe.tcpSocket.port | string | `"http"` |  |
+| readinessProbe.timeoutSeconds | int | `3` |  |
+| readinessProbe.type | string | `"httpGet"` | Probe type: 'httpGet' or 'tcpSocket' |
 | replicaCount | int | `3` | Number of replicas for the Typesense deployment |
-| resources | object | `{}` | Resource requests and limits for the Typesense container |
+| resources | object | `{"requests":{"cpu":"2000m"}}` | Resource requests and limits for the Typesense container. Typesense requires at least 2 vCPUs to operate correctly. No CPU limit is set by default to avoid throttling during indexing and Raft catch-up. |
 | secrets.externalSecret.enabled | bool | `false` | Enable or disable ExternalSecret creation (requires external-secrets operator) |
 | secrets.externalSecret.extractKey | string | `""` | The key path to extract secrets from |
 | secrets.externalSecret.storeName | string | `""` | The name of the ClusterSecretStore or SecretStore to use |
@@ -274,9 +282,13 @@ storage:
 | serviceAccount.automountServiceAccountToken | bool | `false` | Whether to automount the ServiceAccount token |
 | serviceAccount.create | bool | `true` | Whether to create a ServiceAccount |
 | serviceAccount.name | string | `""` | Name of the ServiceAccount. Defaults to fullname |
-| startupProbe.failureThreshold | int | `10` | Number of failed startup checks before marking the container as unhealthy |
-| startupProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense health for startup |
+| startupProbe.failureThreshold | int | `60` | Number of failed startup checks before marking the container as unhealthy |
+| startupProbe.httpGet.path | string | `"/health"` |  |
+| startupProbe.httpGet.port | string | `"http"` |  |
 | startupProbe.periodSeconds | int | `10` | Period (in seconds) to perform the startup check |
+| startupProbe.tcpSocket.port | string | `"http"` |  |
+| startupProbe.timeoutSeconds | int | `3` |  |
+| startupProbe.type | string | `"httpGet"` | Probe type: 'httpGet' or 'tcpSocket' |
 | storage.className | string | `nil` | Storage class to use for Persistent Volume Claims (PVC) |
 | storage.size | string | `"10Gi"` | Size of the persistent storage volume (e.g., 10Gi) |
 | terminationGracePeriodSeconds | int | `300` | Termination grace period in seconds. Typesense recommends 300s to allow graceful shutdown. |

templates/_helpers.tpl

@@ -64,6 +64,31 @@ Create the name of the service account to use
 {{- end }}
 {{- end }}
 
+{{/*
+Render a probe with a single handler selected by .type (httpGet or tcpSocket).
+This avoids Helm's map-merge producing both handlers simultaneously when users
+override just the type in their values, which Kubernetes rejects.
+*/}}
+{{- define "typesense.probe" -}}
+{{- $p := . -}}
+{{- if eq $p.type "tcpSocket" }}
+tcpSocket: {{- toYaml $p.tcpSocket | nindent 2 }}
+{{- else }}
+httpGet: {{- toYaml $p.httpGet | nindent 2 }}
+{{- end }}
+failureThreshold: {{ $p.failureThreshold }}
+periodSeconds: {{ $p.periodSeconds }}
+{{- with $p.timeoutSeconds }}
+timeoutSeconds: {{ . }}
+{{- end }}
+{{- with $p.successThreshold }}
+successThreshold: {{ . }}
+{{- end }}
+{{- with $p.initialDelaySeconds }}
+initialDelaySeconds: {{ . }}
+{{- end }}
+{{- end }}
+
 {{/*
 Create the nodeslist
 */}}

templates/statefulset.yaml

@@ -118,9 +118,9 @@ spec:
               containerPort: {{ .Values.service.port }}
             - name: tcp-peering
               containerPort: 8107
-          startupProbe: {{- toYaml .Values.startupProbe | nindent 12 }}
-          livenessProbe: {{- toYaml .Values.livenessProbe | nindent 12 }}
-          readinessProbe: {{- toYaml .Values.readinessProbe | nindent 12 }}
+          startupProbe: {{- include "typesense.probe" .Values.startupProbe | nindent 12 }}
+          livenessProbe: {{- include "typesense.probe" .Values.livenessProbe | nindent 12 }}
+          readinessProbe: {{- include "typesense.probe" .Values.readinessProbe | nindent 12 }}
           resources: {{- toYaml .Values.resources | nindent 12 }}
           volumeMounts:
           - name: nodeslist

tests/statefulset_test.yaml

@@ -55,19 +55,20 @@ tests:
       - equal:
           path: .spec.template.spec.containers[0].startupProbe
           value:
-            failureThreshold: 10
+            failureThreshold: 60
             httpGet:
               path: /health
               port: http
             periodSeconds: 10
+            timeoutSeconds: 3
       - equal:
           path: .spec.template.spec.containers[0].livenessProbe
           value:
-            failureThreshold: 2
-            httpGet:
-              path: /health
+            failureThreshold: 6
+            tcpSocket:
               port: http
-            periodSeconds: 10
+            periodSeconds: 20
+            timeoutSeconds: 3
       - notExists:
           path: .spec.volumeClaimTemplates[0].spec.storageClassName
       - equal:
@@ -287,11 +288,12 @@ tests:
       - equal:
           path: .spec.template.spec.containers[0].readinessProbe
           value:
-            failureThreshold: 3
+            failureThreshold: 12
             httpGet:
               path: /health
               port: http
-            periodSeconds: 5
+            periodSeconds: 10
+            timeoutSeconds: 3
   - it: should render default anti-affinity but not tolerations or topologySpreadConstraints when empty
     asserts:
       - notFailedTemplate: {}

values.md

@@ -1,6 +1,6 @@
 # typesense
 
-![Version: 1.1.0](https://img.shields.io/badge/Version-1.1.0-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: 30.1](https://img.shields.io/badge/AppVersion-30.1-informational?style=flat-square)
+![Version: 1.1.1](https://img.shields.io/badge/Version-1.1.1-informational?style=flat-square) ![Type: application](https://img.shields.io/badge/Type-application-informational?style=flat-square) ![AppVersion: 30.1](https://img.shields.io/badge/AppVersion-30.1-informational?style=flat-square)
 
 [![Artifact Hub](https://img.shields.io/endpoint?url=https://artifacthub.io/badge/repository/typesense)](https://artifacthub.io/packages/search?repo=typesense)
 
@@ -233,9 +233,13 @@ storage:
 | ingress.enabled | bool | `false` | Enable or disable Ingress for the application |
 | ingress.hosts | list | `[]` | List of hostnames the Ingress will route traffic for |
 | ingress.prefix | string | `"/"` | The URL path prefix for the application |
-| livenessProbe.failureThreshold | int | `2` | Number of failed liveness checks before restarting the container |
-| livenessProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense health for liveness |
-| livenessProbe.periodSeconds | int | `10` | Period (in seconds) to perform the liveness check |
+| livenessProbe.failureThreshold | int | `6` | Number of failed liveness checks before restarting the container |
+| livenessProbe.httpGet.path | string | `"/health"` |  |
+| livenessProbe.httpGet.port | string | `"http"` |  |
+| livenessProbe.periodSeconds | int | `20` | Period (in seconds) to perform the liveness check |
+| livenessProbe.tcpSocket.port | string | `"http"` |  |
+| livenessProbe.timeoutSeconds | int | `3` |  |
+| livenessProbe.type | string | `"tcpSocket"` | Probe type: 'tcpSocket' or 'httpGet'. Default is tcpSocket to check only that the process is alive and listening. Avoid httpGet /health for liveness: Typesense returns 503 when the write queue exceeds --healthy-write-lag (default 500), which occurs normally during Raft catch-up after a restart, causing a liveness-triggered restart loop that prevents the cluster from ever recovering. |
 | metrics.enabled | bool | `false` | Enable Prometheus metrics sidecar |
 | metrics.image.pullPolicy | string | `"IfNotPresent"` | Image pull policy for metrics exporter |
 | metrics.image.repository | string | `"imatefx/typesense-prometheus-exporter"` | Metrics exporter image repository |
@@ -255,11 +259,15 @@ storage:
 | podSecurityContext.runAsGroup | int | `3000` | Group ID for running the Typesense process |
 | podSecurityContext.runAsNonRoot | bool | `true` | Ensure the container does not run as root |
 | podSecurityContext.runAsUser | int | `10000` | User ID for running the Typesense process |
-| readinessProbe.failureThreshold | int | `3` | Number of failed readiness checks before marking the pod as unready |
-| readinessProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense readiness |
-| readinessProbe.periodSeconds | int | `5` | Period (in seconds) to perform the readiness check |
+| readinessProbe.failureThreshold | int | `12` | Number of failed readiness checks before marking the pod as unready |
+| readinessProbe.httpGet.path | string | `"/health"` |  |
+| readinessProbe.httpGet.port | string | `"http"` |  |
+| readinessProbe.periodSeconds | int | `10` | Period (in seconds) to perform the readiness check |
+| readinessProbe.tcpSocket.port | string | `"http"` |  |
+| readinessProbe.timeoutSeconds | int | `3` |  |
+| readinessProbe.type | string | `"httpGet"` | Probe type: 'httpGet' or 'tcpSocket' |
 | replicaCount | int | `3` | Number of replicas for the Typesense deployment |
-| resources | object | `{}` | Resource requests and limits for the Typesense container |
+| resources | object | `{"requests":{"cpu":"2000m"}}` | Resource requests and limits for the Typesense container. Typesense requires at least 2 vCPUs to operate correctly. No CPU limit is set by default to avoid throttling during indexing and Raft catch-up. |
 | secrets.externalSecret.enabled | bool | `false` | Enable or disable ExternalSecret creation (requires external-secrets operator) |
 | secrets.externalSecret.extractKey | string | `""` | The key path to extract secrets from |
 | secrets.externalSecret.storeName | string | `""` | The name of the ClusterSecretStore or SecretStore to use |
@@ -274,9 +282,13 @@ storage:
 | serviceAccount.automountServiceAccountToken | bool | `false` | Whether to automount the ServiceAccount token |
 | serviceAccount.create | bool | `true` | Whether to create a ServiceAccount |
 | serviceAccount.name | string | `""` | Name of the ServiceAccount. Defaults to fullname |
-| startupProbe.failureThreshold | int | `10` | Number of failed startup checks before marking the container as unhealthy |
-| startupProbe.httpGet | object | `{"path":"/health","port":"http"}` | HTTP GET path and port to check Typesense health for startup |
+| startupProbe.failureThreshold | int | `60` | Number of failed startup checks before marking the container as unhealthy |
+| startupProbe.httpGet.path | string | `"/health"` |  |
+| startupProbe.httpGet.port | string | `"http"` |  |
 | startupProbe.periodSeconds | int | `10` | Period (in seconds) to perform the startup check |
+| startupProbe.tcpSocket.port | string | `"http"` |  |
+| startupProbe.timeoutSeconds | int | `3` |  |
+| startupProbe.type | string | `"httpGet"` | Probe type: 'httpGet' or 'tcpSocket' |
 | storage.className | string | `nil` | Storage class to use for Persistent Volume Claims (PVC) |
 | storage.size | string | `"10Gi"` | Size of the persistent storage volume (e.g., 10Gi) |
 | terminationGracePeriodSeconds | int | `300` | Termination grace period in seconds. Typesense recommends 300s to allow graceful shutdown. |

values.yaml

@@ -119,38 +119,59 @@ gateway:
       namespace: "istio-system"
     extras: []
 
-# -- Resource requests and limits for the Typesense container
-resources: {}
+# -- Resource requests and limits for the Typesense container.
+# Typesense requires at least 2 vCPUs to operate correctly. No CPU limit is
+# set by default to avoid throttling during indexing and Raft catch-up.
+resources:
+  requests:
+    cpu: 2000m
 
 livenessProbe:
-  # -- HTTP GET path and port to check Typesense health for liveness
+  # -- Probe type: 'tcpSocket' or 'httpGet'.
+  # Default is tcpSocket to check only that the process is alive and listening.
+  # Avoid httpGet /health for liveness: Typesense returns 503 when the write
+  # queue exceeds --healthy-write-lag (default 500), which occurs normally
+  # during Raft catch-up after a restart, causing a liveness-triggered restart
+  # loop that prevents the cluster from ever recovering.
+  type: tcpSocket
   httpGet:
     path: /health
     port: http
+  tcpSocket:
+    port: http
   # -- Number of failed liveness checks before restarting the container
-  failureThreshold: 2
+  failureThreshold: 6
   # -- Period (in seconds) to perform the liveness check
-  periodSeconds: 10
+  periodSeconds: 20
+  timeoutSeconds: 3
 
 startupProbe:
-  # -- HTTP GET path and port to check Typesense health for startup
+  # -- Probe type: 'httpGet' or 'tcpSocket'
+  type: httpGet
   httpGet:
     path: /health
     port: http
+  tcpSocket:
+    port: http
   # -- Number of failed startup checks before marking the container as unhealthy
-  failureThreshold: 10
+  failureThreshold: 60
   # -- Period (in seconds) to perform the startup check
   periodSeconds: 10
+  timeoutSeconds: 3
 
 readinessProbe:
-  # -- HTTP GET path and port to check Typesense readiness
+  # -- Probe type: 'httpGet' or 'tcpSocket'
+  type: httpGet
   httpGet:
     path: /health
     port: http
+  tcpSocket:
+    port: http
   # -- Period (in seconds) to perform the readiness check
-  periodSeconds: 5
+  periodSeconds: 10
   # -- Number of failed readiness checks before marking the pod as unready
-  failureThreshold: 3
+  failureThreshold: 12
+  timeoutSeconds: 3
 
 # -- Node selector to schedule pods on specific nodes (optional)
 nodeSelector: {}