From eefb55e4f95af5cde1049153656fc6bc4db1b85b Mon Sep 17 00:00:00 2001 From: Guillaume Da Silva Date: Thu, 9 Jul 2026 12:15:31 +0200 Subject: [PATCH] document disabling the CPU limit via resources.override.limit.cpu_in_milli = -1 (QOV-2075) Cross-reference the new -1 sentinel from allow_service_cpu_overcommit. --- docs/configuration/cluster-advanced-settings.mdx | 2 +- docs/configuration/service-advanced-settings.mdx | 10 ++++++++-- 2 files changed, 9 insertions(+), 3 deletions(-) diff --git a/docs/configuration/cluster-advanced-settings.mdx b/docs/configuration/cluster-advanced-settings.mdx index f87d758..9109899 100644 --- a/docs/configuration/cluster-advanced-settings.mdx +++ b/docs/configuration/cluster-advanced-settings.mdx @@ -1996,7 +1996,7 @@ Envoy Gateway automatically selects the best compression algorithm based on the CPU overcommit can cause CPU throttling and unpredictable latency when nodes are under pressure. Use with caution in production. -**Use Case:** Once enabled, you can update the service advanced setting `resources.override.limit.cpu_in_milli` to set a CPU limit higher than the request. +**Use Case:** Once enabled, you can update the service advanced setting `resources.override.limit.cpu_in_milli` to set a CPU limit higher than the request, or set it to `-1` to disable the CPU limit for that service entirely. **Default Value:** `false` diff --git a/docs/configuration/service-advanced-settings.mdx b/docs/configuration/service-advanced-settings.mdx index bfdd75a..fdb43e3 100644 --- a/docs/configuration/service-advanced-settings.mdx +++ b/docs/configuration/service-advanced-settings.mdx @@ -1485,9 +1485,15 @@ Setting `ndots=2` reduces this to **1 DNS query** since the hostname already has **Type:** `integer` -**Description:** Allows you to override the CPU limit for your service pods in millicores (1000m = 1 vCPU). +**Description:** Allows you to override the CPU limit for your service pods in millicores (1000m = 1 vCPU). Set to `-1` to disable the CPU limit entirely, so the container is not capped and can use all the CPU available on the node. -**Use Case:** Use this to fine-tune CPU allocation beyond the standard Qovery instance sizes. +**Use Case:** Use a positive value to fine-tune CPU allocation beyond the standard Qovery instance sizes, or use `-1` to remove CPU throttling for CPU-bound workloads that need to burst freely. + + +Setting a custom CPU limit (including `-1` to disable it) requires CPU overcommit to be enabled on the cluster via the `allow_service_cpu_overcommit` cluster advanced setting. This only affects the CPU limit; the RAM limit (`resources.override.limit.ram_in_mib`) remains mandatory and cannot be disabled, since it protects the pod from being OOMKilled. + + +**Valid values:** `-1` (disabled) or a positive integer greater than the service's CPU request **Default Value:** `null`