fix(helm): podLabel + improve prometheus rules and logging

Change-Id: I3251ea6d7f47d7956db16ec26e13c95fc2d7f23b Signed-off-by: Martin Müller <[email protected]>
CollaboraOnline · Jun 13, 2024 · 685a8a4 · 685a8a4
1 parent bada017
commit 685a8a4
Show file tree

Hide file tree

Showing 6 changed files with 26 additions and 15 deletions.
diff --git a/kubernetes/helm/collabora-online/Chart.yaml b/kubernetes/helm/collabora-online/Chart.yaml
@@ -4,7 +4,7 @@ type: "application"
 name: collabora-online
 description: Collabora Online helm chart
 
-version: 1.1.16
+version: 1.1.17
 appVersion: "24.04.3.1.1"
 
 home: "https://www.collaboraoffice.com/code/"

diff --git a/kubernetes/helm/collabora-online/templates/deployment.yaml b/kubernetes/helm/collabora-online/templates/deployment.yaml
@@ -33,6 +33,9 @@ spec:
       labels:
         {{- include "collabora-online.selectorLabels" . | nindent 8 }}
         type: main
+        {{- with .Values.podLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
     spec:
       {{- with .Values.deployment.hostAliases }}
       hostAliases:

diff --git a/kubernetes/helm/collabora-online/templates/dynamicConfig/flow_dynconfig.yaml b/kubernetes/helm/collabora-online/templates/dynamicConfig/flow_dynconfig.yaml
@@ -19,6 +19,7 @@ spec:
     - parser:
         hash_value_field: "nginx"
         reserve_data: true
+        reserve_time: true
         remove_key_name_field: true
         parse:
           type: "multi_format"
@@ -53,9 +54,8 @@ spec:
             # TODO split ip and port
             nginx.ingress_controller.upstream.address.merged: '${ [record["upstream_address_list"]].flatten&.last }'
             http.response.body.bytes: '${ record["nginx"]["size"] }'
-            event.created: '${ time * 1000 }'
-            event.kind: "event"
             event.category: "web"
+            event.kind: "event"
             event.type: "access"
             event.outcome: '${ record["nginx"]["code"].to_i < 400 ? "success" : "failure" }'
             # for dashboard

diff --git a/kubernetes/helm/collabora-online/templates/prometheus-rules.yaml b/kubernetes/helm/collabora-online/templates/prometheus-rules.yaml
@@ -58,7 +58,7 @@ spec:
       {{- end }}
       {{- range $key, $value := .Values.prometheus.rules.defaults.viewers.doc }}
       - alert: "Collabora Viewers by Document"
-        expr: 'doc_views_active  * on(pid, namespace, job, service, pod, container, endpoint, instance) group_left(key, host, filename) doc_info > {{ $value }}'
+        expr: 'doc_views_active  * on(pid,namespace,job,service,pod,container,endpoint,instance) group_left(key,host,filename) doc_info > {{ $value }}'
         for: "1m"
         labels:
           severity: "{{ $key }}"
@@ -87,7 +87,7 @@ spec:
           summary: '{{ printf "%.0f" $value }}% of all documents are opened simultaneously on different pods in namespace {{ $labels.namespace }}. Viewers can not see each others.'
         `}}
       - alert: "Collabora DocumentsOpenSimultaneously"
-        expr: 'count(doc_info) by (key, namespace, filename, host) > 1'
+        expr: 'count(doc_info) by (key, namespace, host, filename) > 1'
         labels:
           severity: "warning"
         {{`
@@ -129,7 +129,7 @@ spec:
           summary: "we returned an HTTP bad argument to a caller in namespace {{ $labels.namespace }} on pod {{ $labels.pod }}."
         `}}
       - alert: "Collabora Error UnauthorizedRequest"
-        expr: 'increase(error_unauthorized_request[1m]) > 0'
+        expr: 'increase(error_unauthorized_request[{{- .Values.prometheus.rules.defaults.errorUnauthorizedRequest.observationInterval }}]) > {{- .Values.prometheus.rules.defaults.errorUnauthorizedRequest.eventCounter }}'
         labels:
           severity: "warning"
         {{`
@@ -138,7 +138,7 @@ spec:
         `}}
       {{- range $key, $value := .Values.prometheus.rules.defaults.errorServiceUnavailable }}
       - alert: "Collabora Error ServiceUnavailable"
-        expr: 'increase(error_service_unavailable[1m]) > {{ $value }}'
+        expr: 'increase(error_service_unavailable[30m]) > {{ $value }}'
         labels:
           severity: "{{ $key }}"
         {{`

diff --git a/kubernetes/helm/collabora-online/templates/statefulset.yaml b/kubernetes/helm/collabora-online/templates/statefulset.yaml
@@ -26,6 +26,9 @@ spec:
       labels:
         {{- include "collabora-online.selectorLabels" . | nindent 8 }}
         type: main
+        {{- with .Values.podLabels }}
+        {{- toYaml . | nindent 8 }}
+        {{- end }}
     spec:
       terminationGracePeriodSeconds: {{ .Values.terminationGracePeriodSeconds }}
       {{- with .Values.deployment.hostAliases }}

diff --git a/kubernetes/helm/collabora-online/values.yaml b/kubernetes/helm/collabora-online/values.yaml
@@ -30,12 +30,12 @@ collabora:
   aliasgroups: []
 
   extra_params: --o:ssl.enable=false
-  
-  # External hostname:port of the server running coolwsd. 
-  # If empty, it's derived from the request (please set it if this doesn't work). 
+
+  # External hostname:port of the server running coolwsd.
+  # If empty, it's derived from the request (please set it if this doesn't work).
   # May be specified when behind a reverse-proxy or when the hostname is not reachable directly.
   server_name: null
-  
+
   existingSecret:
     enabled: false
     secretName: ""
@@ -68,6 +68,9 @@ prometheus:
         critical: 50
         warning: 2
         info: 0
+      errorUnauthorizedRequest:
+        observationInterval: "60m"
+        eventCounter: 8
       errorStorageConnections:
         critical: 50
         warning: 2
@@ -114,6 +117,8 @@ logging:
 
 podAnnotations: {}
 
+podLabels: {}
+
 podSecurityContext: {}
 #  fsGroup: 2000
 
@@ -237,8 +242,8 @@ dynamicConfig:
     enabled: false
     ecs: false
     dedot:
-    additionalFilters: [ ]
-    localOutputRefs: [ ]
+    additionalFilters: []
+    localOutputRefs: []
     globalOutputRefs:
       - "default"
 
@@ -288,8 +293,8 @@ dynamicConfig:
       enabled: false
       ecs: false
       dedot:
-      additionalFilters: [ ]
-      localOutputRefs: [ ]
+      additionalFilters: []
+      localOutputRefs: []
       globalOutputRefs:
         - "default"