Skip to content
SYNNET DevHub

Bugs / findings log — night 2026-06-11

Synced read-only from /home/synnet/mirrors/synvirt-product/docs/night/20260611/BUGS-FOUND.md. Edit at the source, not here.

Bugs / findings log — night 2026-06-11

Section titled “Bugs / findings log — night 2026-06-11”

Running list; items graduate to fixes (with regression tests) or to the morning report.

1. /api/v1/host/usage is unauthenticated

Section titled “1. /api/v1/host/usage is unauthenticated”
  • Found during Phase 1 profiling (it was the only fast endpoint under Basic-Auth load — because it never touches PAM).
  • It sits in the PUBLIC pre-auth router section (web/routes.rs ~:97), between the deliberately-public identity/setup/install routes. Host CPU/RAM/load figures are exposed to anyone who can reach :443.
  • Suspected oversight, BUT the pre-admin setup/installer surfaces might legitimately read it (host sizing display) — verify before moving it behind auth. → Phase 7.

2. Pre-existing ESLint errors (6) on 0.15.0

Section titled “2. Pre-existing ESLint errors (6) on 0.15.0”
  • VmDetailHero (unused isStopped), CdromCard (unused Plus), NicCard (parsing error: unexpected null character!), PciPassthroughCard (unused h), HomeView (unused Disc3, Database).
  • NicCard’s null character is the alarming one — a stray \0 in source. → Phase 7 mechanical fixes.

3. Pre-existing vitest failures (5) on 0.15.0

Section titled “3. Pre-existing vitest failures (5) on 0.15.0”
  • isoLibrary deleteIso spec (API gained force arg, spec stale), IsoActionsMenu, Button primary variant, WizardStepper error state, Tag severity classes. Match the known baseline. → Phase 7 if time.

4. pgrep -x qemu-system-x86_64 never matches

Section titled “4. pgrep -x qemu-system-x86_64 never matches”
  • comm is truncated to 15 chars; the exact name is 18. Guest-PID checks silently passed vacuously (empty == empty). Canonical pattern is pgrep -f '[q]emu-system' (bracket defeats self-match). Fixed in the bench harness + fleet deploy script; grep the repo for other pgrep -x qemu instances in Phase 7.

5. /etc/synvirt/release stale after hot deploys (known gotcha)

Section titled “5. /etc/synvirt/release stale after hot deploys (known gotcha)”
  • All three nodes report Fix45-54 stamps while actually running Fix77 builds. Known; the updater owns the stamp. Candidate for stage_apply to rewrite the stamp on hot deploys → Phase 7 discussion item.

6. .65 synvirt-network.service failing (host state, pre-existing)

Section titled “6. .65 synvirt-network.service failing (host state, pre-existing)”

journalctl on .65/synvirt-03: no bridge named ens192 + /etc/systemd/resolved.conf.d: No such file or directory namespace failures. .65 was reinstalled with NO OVS (RECON noted it); the boot network service can’t reconcile a bridge that was never created. NOT a tonight regression — NFS/iSCSI (round 3) don’t need OVS and passed. Morning: run the New vSwitch wizard on .65 to build vSwitch0, or re-run the installer’s network seed. Minor: synvirt-network.service should mkdir -p /etc/systemd/resolved.conf.d in ExecStartPre so the namespace mount doesn’t fail on a fresh host.

7. .75 clock skew (+5 months) breaks apt

Section titled “7. .75 clock skew (+5 months) breaks apt”

nfs-common has “no installation candidate” on .75 because the host clock is ~5 months ahead → repo signatures read as not-yet-valid. Known (.75 NTP UDP123 blocked). Blocks NFS on .75 → blocks the cross-host Guardian drill. Host prereq.