Omni always look into image factory despite i put machine configs patch.

[edyzakaria]

Hi,

I have been knocking my head due to issue that my host machine can't reach Azure repository and got "connection reset by peer" then followed by "connection refused".
I was having this issue with talos and successfully resolve it using talos guide on air-gapped as referred in this link:
https://www.talos.dev/v1.9/advanced/air-gapped/
Through that guide i even able to download the 1.9.3 image and 1.32.1 kube and successfully installed it. I am so happy since previously i also knocking my head due to the issue Azure and my provider (in indonesia).

Then i bring the same things to omni. It failed. it always looking to image factory in Azure, despite i apply the config patch in cluster, machine, or in the control plane gear icon. I really confused.

Then i also doing the guide for omni airgapped using gitea. so i pushed the images also to gitea based on this link:
https://omni.siderolabs.com/tutorials/install-airgapped-omni

Do the patch also, and i still get that talos always looking on the Azure despite i have put the config patch either single place or any place.

In the talos installation i can easily use the patch and checked on the controlplane.yaml to ensure it and get understanding better (learned the hard way).

But in omni, i really confused, i would like to manually edit the controlplane.yaml, but i dont find way into it.

Edit: Below is my docker compose content:

version: '3.8'
services:
  omni:
    container_name: omni
    restart: unless-stopped
    image: "ghcr.io/siderolabs/omni:${OMNI_IMG_TAG}"
    devices:
      - /dev/net/tun
    volumes:
      - ${ETCD_VOLUME_PATH}:/_out/etcd
      - ${ETCD_ENCRYPTION_KEY}:/omni.asc
      - ${TLS_CERT}:/tls.crt
      - ${TLS_KEY}:/tls.key
    network_mode: "host"
    cap_add:
      - NET_ADMIN
    command: >
      --account-id=${OMNI_ACCOUNT_UUID} 
      --cert=/tls.crt
      --key=/tls.key
      --machine-api-cert=/tls.crt
      --machine-api-key=/tls.key
      --private-key-source='file:///omni.asc' 
      --event-sink-port=${EVENT_SINK_PORT}
      --bind-addr=${BIND_ADDR}
      --machine-api-bind-addr=${MACHINE_API_BIND_ADDR}
      --k8s-proxy-bind-addr=${K8S_PROXY_BIND_ADDR}
      --advertised-api-url=${ADVERTISED_API_URL}
      --advertised-kubernetes-proxy-url=${ADVERTISED_K8S_PROXY_URL}
      --siderolink-api-advertised-url=${SIDEROLINK_ADVERTISED_API_URL}
      --siderolink-wireguard-advertised-addr=${SIDEROLINK_WIREGUARD_ADVERTISED_ADDR}
      --initial-users=${INITIAL_USER_EMAILS}
      --talos-installer-registry=${GITEA_HOSTNAME}:3000/siderolabs/installer 
      --kubernetes-registry=${GITEA_HOSTNAME}:3000/siderolabs/kubelet 
      ${AUTH}

Is there anything lack?
In the guide there is '--talos-imager-image' flag which failed to load.
My environment is not air-gapped. it just simple homelab for learning.
I can pull docker image even omni, gitea, talos, no issue.
Only when it loaded into proxmox vm it hit the issue.

Is it always like that?
No. Because i try the non-plugin talos image 1.9.2 and it is success. As long as didnt put the flag:
--install-image
Then i successfully build a cluster of 3 control planes and 2 workers.

Please help.

[edyzakaria]

Solved it already. I override through config patch for the issue.