From fa9d0e9cbeb681d990dea69da4a24205e7028455 Mon Sep 17 00:00:00 2001 From: Christos Alexiou Date: Tue, 19 May 2026 23:52:55 +0300 Subject: [PATCH] fix: fine tune ops timing values --- charts/budibase/values.yaml | 57 +++++++++++++++++++++++++------------ 1 file changed, 39 insertions(+), 18 deletions(-) diff --git a/charts/budibase/values.yaml b/charts/budibase/values.yaml index ca5aa9c56c..360c2719dc 100644 --- a/charts/budibase/values.yaml +++ b/charts/budibase/values.yaml @@ -178,9 +178,9 @@ services: couchdb: "http://{{ .Release.Name }}-svc-couchdb:{{ .Values.services.couchdb.port }}" # -- The time in seconds to wait before sending SIGKILL after SIGTERM. # Must be greater than preStopDelaySeconds to allow graceful shutdown. - terminationGracePeriodSeconds: 75 + terminationGracePeriodSeconds: 70 # -- Seconds to sleep before SIGTERM to allow load balancer deregistration. - preStopDelaySeconds: 40 + preStopDelaySeconds: 45 # -- Pod readiness gates for the proxy pods. Useful for ALB target group # health tracking with the AWS Load Balancer Controller. # Example: [{conditionType: "target-health.elbv2.k8s.aws/my-tgb"}] @@ -204,9 +204,11 @@ services: port: 10000 scheme: HTTP # @ignore - failureThreshold: 40 + failureThreshold: 24 # @ignore - periodSeconds: 3 + periodSeconds: 5 + # @ignore + timeoutSeconds: 1 # -- Readiness probe configuration for proxy pods. You shouldn't need to # change this, but if you want to you can find more information here: # @@ -275,12 +277,12 @@ services: # -- The amount of time to wait between requesting a shutdown and killing the # container. This is used to give the apps service time to finish processing # any requests before shutting down. You shouldn't need to change this. - terminationGracePeriodSeconds: 75 + terminationGracePeriodSeconds: 70 # -- Seconds to wait after pod is Ready before adding to endpoints. # Ensures pods are stable before receiving traffic. minReadySeconds: 10 # -- Seconds to sleep before SIGTERM to allow load balancer deregistration. - preStopDelaySeconds: 40 + preStopDelaySeconds: 45 # -- Pod readiness gates for the apps pods. Useful for ALB target group # health tracking with the AWS Load Balancer Controller. # Example: [{conditionType: "target-health.elbv2.k8s.aws/my-tgb"}] @@ -310,9 +312,11 @@ services: port: 4002 scheme: HTTP # @ignore - failureThreshold: 30 + failureThreshold: 24 # @ignore - periodSeconds: 3 + periodSeconds: 5 + # @ignore + timeoutSeconds: 1 # -- Readiness probe configuration for apps pods. You shouldn't need to # change this, but if you want to you can find more information here: # @@ -383,9 +387,9 @@ services: # the container. This is used to give the automation worker service time to # finish processing any requests before shutting down. You shouldn't need to # change this. - terminationGracePeriodSeconds: 75 + terminationGracePeriodSeconds: 70 # -- Seconds to sleep before SIGTERM to allow load balancer deregistration. - preStopDelaySeconds: 40 + preStopDelaySeconds: 45 # -- Pod readiness gates for the automation worker pods. Useful for ALB # target group health tracking with the AWS Load Balancer Controller. # Example: [{conditionType: "target-health.elbv2.k8s.aws/my-tgb"}] @@ -416,9 +420,11 @@ services: port: 4002 scheme: HTTP # @ignore - failureThreshold: 30 + failureThreshold: 12 # @ignore - periodSeconds: 3 + periodSeconds: 30 + # @ignore + timeoutSeconds: 1 # -- Readiness probe configuration for automation worker pods. You shouldn't # need to change this, but if you want to you can find more information # here: @@ -490,12 +496,12 @@ services: # the container. This is used to give the worker service time to finish # processing any requests before shutting down. You shouldn't need to change # this. - terminationGracePeriodSeconds: 75 + terminationGracePeriodSeconds: 70 # -- Seconds to wait after pod is Ready before adding to endpoints. # Ensures pods are stable before receiving traffic. minReadySeconds: 10 # -- Seconds to sleep before SIGTERM to allow load balancer deregistration. - preStopDelaySeconds: 40 + preStopDelaySeconds: 45 # -- Pod readiness gates for the worker pods. Useful for ALB target group # health tracking with the AWS Load Balancer Controller. # Example: [{conditionType: "target-health.elbv2.k8s.aws/my-tgb"}] @@ -525,9 +531,11 @@ services: port: 4003 scheme: HTTP # @ignore - failureThreshold: 40 + failureThreshold: 24 # @ignore - periodSeconds: 3 + periodSeconds: 5 + # @ignore + timeoutSeconds: 1 # -- Readiness probe configuration for worker pods. You shouldn't need to # change this, but if you want to you can find more information here: # @@ -658,9 +666,9 @@ services: # -- Number of litellm replicas. replicaCount: 2 # -- The time in seconds to wait before sending SIGKILL after SIGTERM. - terminationGracePeriodSeconds: 75 + terminationGracePeriodSeconds: 70 # -- Seconds to sleep before SIGTERM to allow load balancer deregistration. - preStopDelaySeconds: 40 + preStopDelaySeconds: 45 # -- Pod readiness gates for the litellm pods. Useful for ALB target group # health tracking with the AWS Load Balancer Controller. # Example: [{conditionType: "target-health.elbv2.k8s.aws/my-tgb"}] @@ -669,6 +677,19 @@ services: # Requires podDisruptionBudget.enabled to be true. # Set minAvailable to override the global podDisruptionBudget.minAvailable for this service. pdb: {} + # -- Startup probe configuration for litellm pods. Uses a TCP socket check + # because litellm doesn't expose a dedicated /health endpoint on startup. + # @default -- TCP socket check on the litellm port. + startupProbe: + # @ignore + tcpSocket: + port: 4000 + # @ignore + failureThreshold: 24 + # @ignore + periodSeconds: 5 + # @ignore + timeoutSeconds: 1 # -- Resources for litellm pods. resources: {} # -- Extra environment variables for the litellm container.