HydraIssues

Issues / #761
Bind the mobile-kit workshop fleet into one district
done improvement Project: hydracluster Reporter: cederik 1 Oct 2026 20:58

Description

Workshop on 2026-10-02 uses the mobile kit as the venue. Students self-enroll their own laptops at /enroll; the demo spine is five devices the owner brings.

Problem: the five devices sit in THREE districts, and body selection is per district, so nothing can stream across them.

  • spicy-cactus-76 (MSI1060, body, Omarchy, wivrn) was bxl1-test / ad6
  • chunky-turnip-23 (body, Windows, alvr+hydrabrain) was bxl1-test-2 / cloud-seven
  • quest-head-36afcd (hydraheadquest v0.8.2) was bxl1-test-2 / cloud-seven
  • ipad-head-cederik (hydraheadipad v0.2.158) was bxl1 / cloud-seven
  • cranky-toaster-86 (hydraheadflatscreen 9.9.9-m4test) was bxl1-test-2 / cloud-seven

Action: set all five to district bxl1-test-2, venue mobile-kit. Owner ruled bxl1-test-2 (keeps the set off the production district). hydrabrain on chunky stays enabled; no heavy apps expected.

Residuals found while checking, each needs its own ticket or decision:

  1. The laptop-body path is not in a production hydrabody release. sunshine_startup_linux.go first appears in v2.0.70-rc.10 and wivrn_linux.go in v2.0.70-rc.13. There is no v2.0.70 final tag. Production is v2.0.69, whose provision_linux.go is an 874-byte stub. Both bodies therefore stay pinned to release_channel staging, and they are on different RCs (MSI rc.13, chunky rc.9).
  2. hydrabody-linux.yaml hard-exits without pacman (Arch/Omarchy only), so a student laptop on any other distro cannot become a body.
  3. The venue registry still records mobile-kit under district bxl1 and has no entry for bxl1-test-2. Node records are not validated against it, so the fleet works, but the admin district form will not offer mobile-kit. Needs PUT /api/v1/venues/mobile-kit with district_id bxl1-test-2, which requires the hydravenues token.
  4. cranky-toaster-86 runs a sideloaded 9.9.9-m4test against hydrahead production 2.2.5. Because 9.9.9 sorts higher the updater never fires, which is #534. Its #500 AppImage fix was republished at the same version, so it may still hold the build that crashes about 11s into a stream.
  5. ipad-head-cederik last reported wireguard disconnected.
  6. Body double-booking (#663) is still open. Three heads and two bodies means heads must be assigned explicitly, not left to pick.
  7. hydraheadwebstream.experiencenet.com does not answer and its host grumpy-hamster-99 is no longer enrolled; hydraneckwebrtc.experiencenet.com has no route; hydrahead.experiencenet.com has no DNS while still being the server_url in the hydraheadwebstream runbook. Out of scope for this workshop (owner ruled no webstream) but recorded here.

Sub-issues (1)

done #762 Portal links creators to a PRIVATE template repo (404 for everyone outside the org)

Session Context

Venue
mobile-kit
District
bxl1-test-2

Comments (1)

api 2 Oct 2026 17:30

The binding itself is done and verified: all five nodes sit in district bxl1-test-2, venue mobile-kit, and the venue registry now agrees with them (it said bxl1 while the nodes said bxl1-test-2). mobile-kit also moved to the new hydramancers organization, which is what makes it visible to a self-registered creator.

The single most useful thing found here is worth repeating outside this ticket: venue is a hard filter, not a preference. In hydracluster/pkg/api/handlers_api.go the eligible-bodies handler DROPS a body unless body.Venue == head.Venue or the two share a non-empty owner. So district alignment alone is not enough, and any machine that should serve the kit's heads needs the venue set too.

The workshop this was filed for did not run in the shape originally planned, and the residuals listed in the description are all still true and still open:

  • the laptop-body path is still only on hydrabody staging RCs, with no v2.0.70 final
  • hydrabody-linux.yaml still hard-exits without pacman
  • cranky-toaster-86 still runs the 9.9.9-m4test sideload that can never self-update (#534)
  • ipad-head-cederik last reported wireguard disconnected
  • body double-booking is still open under #663
  • the hydraheadwebstream host is still gone and hydrahead.experiencenet.com still has no DNS

Closing this one because the fleet binding it asked for is complete. Those residuals each deserve their own ticket rather than living in a closed one; none of them were addressed.