Nice. One additional recommendation is to change the default SSH port as 22 is the common port for bot to probe or attack.

I hate to say this but sounds like a hard drive failure or corrupt OS. When it happen to me it was my HD. Either case it may require a re-install.

Good write-up. Dependencies can bite you real hard if not careful. Running a Proxmox 3 node cluster with 20 VMs and along with 80 docker containers can be a challenge to maintain so automation has been my lifesaver. Been considering setting up a time server for the homelab since I had a time issue on one of my server.

I use Gitea and run my Ansible playbooks. Currently adding my n8n workflows to Gitea.

@David Vincent To answer both your questions. After installing Ollama, you will need to edit the ollama.service file as root: "systemctl edit ollama.service". Add the following and than restart the service. [Service] Environment="OLLAMA_HOST=0.0.0.0:11434" Environment="OLLAMA_CONTEXT_LENGTH=64000" <---Context Environment="OLLAMA_KV_CACHE_TYPE=q8_0" Environment="OLLAMA_GPU_COUNT=2" Environment="OLLAMA_GPU_MEMORY_LIMIT=24GB" Environment="OLLAMA_GPU_SPLIT=0.5,0.5" <----- Splits the GPUs Environment="OLLAMA_FLASH_ATTENTION=1" Environment="OLLAMA_NUM_CTX=8192" Environment="OLLAMA_MAX_LOADED_MODELS=3 Environment="OLLAMA_KEEP_ALIVE=2" Environment="OLLAMA_NUM_PARALLEL=4" These setting will optimized Ollama for better performance but can be modify to fit your needs.

Also testing the new qwen3.6:35b model which so far has been producing good results.