[ci skip] Fix docker-registry VM: add SSH key, remove hourly restart cron

- Set explicit devvm SSH public key for cloud-init (was empty, breaking SSH access)
- Remove hourly cron that restarted all registry containers, which wiped the
  in-memory blobdescriptor cache and caused low pull-through cache hit rates
This commit is contained in:
Viktor Barzin 2026-02-15 22:16:41 +00:00
parent 6f33c3008f
commit 0eac3d6de6

View file

@ -266,7 +266,7 @@ module "docker-registry-template" {
proxmox_user = "root" # SSH user on Proxmox host
ssh_private_key = var.ssh_private_key
ssh_public_key = var.ssh_public_key
ssh_public_key = "ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIDHLhYDfyx237eJgOGVoJRECpUS95+7rEBS9vacsIxtx devvm"
cloud_image_url = local.cloud_init_image_url
image_path = local.non_k8s_cloud_init_image_path # keke
@ -290,7 +290,8 @@ module "docker-registry-template" {
)
),
"( crontab -l 2>/dev/null; echo '0 3 * * 0 /usr/bin/docker exec registry registry garbage-collect -m /etc/docker/registry/config.yml' ) | crontab -",
"( crontab -l 2>/dev/null; echo '0 * * * * /usr/bin/docker restart registry registry-ghcr registry-quay registry-k8s registry-kyverno' ) | crontab -",
# Hourly restart cron removed - it wiped the in-memory blobdescriptor cache every hour,
# causing low cache hit rates on the pull-through proxy. Docker containers use --restart always.
"docker run -p 5000:5000 -p 5001:5001 -d --restart always --name registry -v /etc/docker-registry/config.yml:/etc/docker/registry/config.yml registry:2",
# ghcr.io proxy
"mkdir -p /etc/docker-registry/ghcr",