Building and running the machine

Why we aren't just hacking.
Building and running the machine

I think this note from Chip (@ChipTuner) is one of those things that people with business management experience take a lot more seriously than most developers and influencers do.

I am painfully aware of the cost of systems administration, financial transaction management and recordkeeping, recruiting and personnel management, legal and compliance, requirements management, technical support, renting and managing physical spaces and infrastructure, negotiating with suppliers, customer service, etc. etc.

There’s this idea, on Nostr, that sort of trickled in along with Bitcoin Twitter, that we would all just be isolated subsistance farmers and one-man-show podcasters with a gigantic server rack in the basement. But some of us are running real companies – on and off Nostr, for-profit and non-profit – and it often requires a lot of human labor.

The things we build aren’t meant to be used by one person and his girlfriend and his dog. Yes, he can also run all these things, himself, but he no longer has to. Our existence gives him the choice: run these things or pay us to run them and spend your time doing something else, that you do better than we do.

These things are meant to be used by hundreds… thousands… eventually millions of people. The workflows, processes, infrastructure, and personnel need to be able to scale up-and-down, scale in-and-out, work smoothly with 5 people or 50 people. These are the sort of Nostr systems that wouldn’t collapse when encountering a sudden influx or mass-escape. But these systems are much more complex and they take time to build and staff to run them. (And, no, AI can’t replace them all. AI means that they now also have to integrate a bunch of AI into the system and maintain that, too.)

GitCitadel (@GitCitadel) is very automation-forward, but we still have to front the incredibly high cost of designing and building the automation, train people to interact with it (there are now over 20 people integrated into the workflow!), adjust it based upon their feedback, and we have to support the automation, once it’s running.

This sort of streamlined machine is what people pay companies for, not code. That is why there’s little business cost to open source.

Open-source is great, but…

meme

ChipTuner
Feb 9, 2025 23:25

Feels good, but sysadmin is so much work sometimes XD

  • secondary file server is back in healthy working order with realtime sync
  • daily snapshots on both servers are working correctly
  • clients can fail-over file services transparently, and it works again
  • encrypt everything where possible
  • all hard drives over 30k hours were retired.
  • hosts, guests, containers, storage, network devices, firmwares are all up to date (huge success)
  • failover/migrations tested (again) and successfully working across all hosts
  • guest applications & containers all up to date (manually check patch notes and update container hashes)
  • all firewalls and rules were audited, updated, and tested
  • all apps were tested to successfully persist reboots (also big accomplishment)
  • fixed some performance issues with L3 routing and cifs shares
  • audited DNS servers and updated blocklists
  • less than 2 minutes downtime for my customer’s services (also big success) would be less if I had redundant networking gear. Maybe in the near future.


Write a comment
No comments yet.