gitadmin/foxhunt
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

infra: H100 pool min_size=1 — keep node warm, skip 15-min driver install

Browse Source 
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) <noreply@anthropic.com>
This commit is contained in:
jgrusewski
2026-04-17 09:12:36 +02:00
parent 0a31bf729a
commit 33e9d80600
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
infra/modules/kapsule/main.tf
View File

@@ -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