Week 51 of 2025: The First Week of Getting Back on Track

One-liner

• This week I reviewed and adjusted my Helm deployment approach, published two HomeLab debugging posts, and made time for friends—while still digesting my thinking on incident workflows and SLA.

What I did

• Work

  • Optimized the deployment strategy of a project's Helm chart. I used to run helm upgrade ... locally for speed, which worked, but it becomes problematic once the project is fully deployed to production.
    • The previous “deploy cluster by cluster” approach makes upgrading the project's image across all clusters painful.
  • A problem
    • I still lack a clear direction on incident handoff/workflow thinking. I have been reading what Ant Group is doing, but I haven't internalized it.

• Tech

• Life

  • On Sunday, I went to Hongqiao Art Center with my beautiful partner to watch a play, "The Lawless Cowboy of the West". Honestly, I didn't like it. I almost fell asleep for ten minutes in the middle. Too many cheesy jokes and era memes.
  • This weekend I met up with good friends for Texas Hold'em, dinner, and a “we almost went to the Bund but ended up on Huaihai Road instead”.
    • Won a bit at Hold'em, lol. Turns out I'm still pretty sharp at 1:20.
    • Ate Quzhou cuisine and Guizhou beef & mutton hotpot with the boys.

Output

• Writing / Notes

  • [x] [Daily Debugging: Minisforum N100 Random Disk Drop / System Freeze (DNS, 1+ Day Uptime)](../minisforum-n100-dns-random-freeze)
  • [x] [Daily Debugging: PVE Cross-VLAN Ping Fails (Wrong Default Gateway + OpenWrt LAN NAT)](../pve-openwrt-icmp-debug)

• Code / Projects

  • (none)

• Reading / Videos

  • (none)

Review

• Keep doing

  • This week I continued working around incident workflow and SLA, but I still need to summarize and internalize more.

• Stop doing

  • I need to better balance learning and entertainment — my weight has been going up again recently.

• One lesson

Next week plan (Week 52)

• Digest Ant Group papers and build my own framework for SLA and incident workflow thinking.
• Digest the vLLM engine, try implementing a few issues, and write a post that explains what it is clearly.
• Improve my AI toolchain and workflow.

Links