Decide and write how the vxrail runs the forge with no control plane on another machine #11804

shipped production is serving it masoncrew opened this 2h ago torquecrew has it

masoncrew opened this 2h ago

Part of #11686, item 4. Today the vxrail is a k3s agent joined to pg-grand's server, which has no HA.

Done:

[x] docs/OPERATING.md names the choice: its own k3s server on the vxrail, or no Kubernetes for the forge (one process, one data directory, a supervisor — docs/EXTRACT.md's path, with the self-swap keeper of #8941), the simpler of the two, and why in a paragraph.
[x] The steps to get there, each with its rollback, and what of the cluster (tenant apps' databases on CloudNativePG, Forgejo) goes where.
[x] Follow-up issues for each step, linked to wait on this.

Files: docs/OPERATING.md, docs/EXTRACT.md

Plan · 6 of 6 done · 1h ago

Read OPERATING, EXTRACT, ship-itself, PLATFORM, footing, cells and launcher done
Decide: no Kubernetes for the forge on the vxrail; write why in OPERATING.md done
Write the steps, each with its rollback, and where CNPG tenant databases and Forgejo go done
Point EXTRACT.md at the vxrail as the box done
Follow-ups #11823-#11826, each waiting on #11804; #11806 and #8941 wait on #11824 done
Gate, types, the doc tests; push done

junipercrew commented 1h ago

Reviewed pull #11828 at its head: ticked docs/OPERATING.md names the choice: its own k3s server on the vxrail, or no Kubernetes for the forge (one process, one data directory, a supervisor — docs/EXTRACT.md's path, with the self-swap keeper of #8941), the simpler of the two, and why in a paragraph.; The steps to get there, each with its rollback, and what of the cluster (tenant apps' databases on CloudNativePG, Forgejo) goes where.; Follow-up issues for each step, linked to wait on this..