[Feature]: Configure kdump and rasdaemon for troubleshooting #144

Closed
opened 2026-08-30 14:33:42 +00:00 by ssmithx · 1 comment
Collaborator

Problem

More data in case of a system lockup

Proposed Solution

see md

Alternatives Considered

No response

Area

Security

### Problem More data in case of a system lockup ### Proposed Solution see md ### Alternatives Considered _No response_ ### Area Security
Collaborator

Implemented and shipped to main — closing per the working rule (done + self-verified at code level; deploy-time acceptance happens with the release).

What shipped:

  • cbd463e9 — feat(host): the host-fixup OTA channel + kdump/rasdaemon (approved decisions: hang capture ON, crashkernel=256M, backfill with the release, phase-2 UI deferred): kdump-tools/kexec-tools/rasdaemon install, USE_KDUMP=1 + /var/crash + compressed core collector, hang/panic sysctl policy, GRUB crashkernel append (next-reboot caveat logged), rasdaemon enablement, keep-2 dump pruning. Unit tests 4/4, cargo fmt clean.
  • 1872fc20 — feat(image): the identical end state baked into fresh ISOs (packages+config+enable in the rootfs build, crashkernel=256M on the installed GRUB cmdline) — edited in the tracked builder (image-recipe/_archived/build-auto-installer-iso.sh), with the generated workspace synced to match.
  • tests/lifecycle/os-audit.sh section D — graded baseline check (FAIL never-landed / WARN awaiting reboot / PASS reserved + policy live + rasdaemon recording), independent of RPC health.
  • docs/kdump-rasdaemon-design.md (marked IMPLEMENTED) + docs/system-level-ota-design.md (the general channel + policy).

Verification still owed before the release TAG, tracked in the release runbook (not here): archi-dev-box takes this OTA build, reboots once so crashkernel actually reserves, then os-audit section D must report PASS as part of the lifecycle gate. One manual crash drill (sysrq c) on the box recommended once, deliberately not in the gate.

Implemented and shipped to main — closing per the working rule (done + self-verified at code level; deploy-time acceptance happens with the release). **What shipped:** - `cbd463e9` — feat(host): the host-fixup OTA channel + kdump/rasdaemon (approved decisions: hang capture ON, crashkernel=256M, backfill with the release, phase-2 UI deferred): kdump-tools/kexec-tools/rasdaemon install, USE_KDUMP=1 + /var/crash + compressed core collector, hang/panic sysctl policy, GRUB crashkernel append (next-reboot caveat logged), rasdaemon enablement, keep-2 dump pruning. Unit tests 4/4, cargo fmt clean. - `1872fc20` — feat(image): the identical end state baked into fresh ISOs (packages+config+enable in the rootfs build, crashkernel=256M on the installed GRUB cmdline) — edited in the tracked builder (image-recipe/_archived/build-auto-installer-iso.sh), with the generated workspace synced to match. - `tests/lifecycle/os-audit.sh` section D — graded baseline check (FAIL never-landed / WARN awaiting reboot / PASS reserved + policy live + rasdaemon recording), independent of RPC health. - docs/kdump-rasdaemon-design.md (marked IMPLEMENTED) + docs/system-level-ota-design.md (the general channel + policy). **Verification still owed before the release TAG, tracked in the release runbook (not here):** archi-dev-box takes this OTA build, reboots once so crashkernel actually reserves, then os-audit section D must report PASS as part of the lifecycle gate. One manual crash drill (sysrq c) on the box recommended once, deliberately not in the gate.
chaum closed this issue 2026-08-31 11:29:22 +00:00
Sign in to join this conversation.
2 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: lfg2025/archy#144