AI Disclosure:
I’ll be using AI for some guidance and some tasks
I need to rebuild my homelab. Projects I want to work on:
- Running Kubernetes at home with Talos. I’ve used EKS and kOps at work.
- Automated patching and building of Arch kernels into a local repository I can pull. I use the kakra patches to control btrfs metadata placement on optane. This will run on Kubernetes.
- Play with Ceph. I never have.
- Patch BIOSes. I recently acquired 10 small Haswell systems that don’t support NVMe booting. I want to use 16 GB optane drives as boot drives. These will be Talos nodes.
- Establish my own software factory. I’m currently using over 10 million tokens per week but it’s all through prompting. I have a couple 7900 XTX that should be able to run mixture of experts models for the easier coding work; I’ll still have frontier models for challenging bugs and architecture.
- I need to build out my local test fleet of GPUs for testing builds of software I help maintain. I have another set of boxen that can host full size (old) GPUs. I need to configure background pods that are automatically evicted when a test run needs a GPU. I’ll still be using Vast and RunPod to access modern Nvidia GPUs.
- I need to create a local backup server out of old 3TB drives I have. I have remote backups but not local backups.
That will probably eat up all my time.
Other projects I’d like to do:
- Put Linux on this old PowerMac G5 I bought so I can do big endian testing without emulating an s390. It’s so old that emulating an s390 is probably faster though!
- Setup a local ADSB platform with browsing of historical data.
- It would be nice to add an M-series Mac for local testing.