From 33e9d80600b4ade0e47380ca005732f42f5d4c18 Mon Sep 17 00:00:00 2001 From: jgrusewski Date: Fri, 17 Apr 2026 09:12:36 +0200 Subject: [PATCH] =?UTF-8?q?infra:=20H100=20pool=20min=5Fsize=3D1=20?= =?UTF-8?q?=E2=80=94=20keep=20node=20warm,=20skip=2015-min=20driver=20inst?= =?UTF-8?q?all?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Scaleway Kapsule GPU nodes need NVIDIA driver DKMS compile on cold start (~15 min). Setting min_size=1 keeps the node warm between runs. Scale to 0 manually when done training for the day. Co-Authored-By: Claude Opus 4.6 (1M context) --- infra/modules/kapsule/main.tf | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/infra/modules/kapsule/main.tf b/infra/modules/kapsule/main.tf index 9e30c54fb..7804f3c7e 100644 --- a/infra/modules/kapsule/main.tf +++ b/infra/modules/kapsule/main.tf @@ -96,7 +96,7 @@ resource "scaleway_k8s_pool" "ci_training_h100" { name = "ci-training-h100" node_type = var.ci_training_h100_type size = 1 - min_size = 0 + min_size = 1 # Keep warm: avoids 15-min NVIDIA driver install on cold start. Scale to 0 when idle. max_size = var.ci_training_h100_max_size autoscaling = true autohealing = true