Troubleshooting

Supported platforms

grain supports macOS and Linux hosts (amd64 / arm64) with hardware virtualization. Native Windows is not a host OS — use the remote API/SDK against a macOS or Linux machine. WSL reports as Linux and follows the same virt requirements as other Linux hosts (see Install → Supported platforms). Nested virtualization only matters when grain itself runs inside another VM without hardware virt exposed to that guest.

grain doctor

grain doctor

Checks:

Hard failures print ✗ and cause a non-zero exit. Soft items print · and do not fail doctor.

Fix reported ✗ lines, then re-run doctor until all good.

Serial and QEMU logs

grain logs              # only VM, or prompts if many
grain logs sbox-1
grain logs -f sbox-1    # follow (tail -F style)
grain logs --qemu sbox-1

Use serial for cloud-init, kernel, and login noise. Use --qemu for hypervisor start failures (missing firmware, bad args, accelerate errors).

Works offline for a named VM by path; without a name it prefers the daemon list, then scans vms/.

VM will not become ready / SSH timeout

Create waits up to the API client timeout (~5 m); SSH readiness uses ready_timeout (default 2m).

1. grain logs -f <name> — look for cloud-init errors or stuck boot.
2. grain logs --qemu <name> — QEMU failed immediately?
3. Confirm image: grain image ls / grain image pull.
4. On Apple Silicon, confirm EDK2 firmware is present (see UEFI below).
5. Increase wait in config:

ready_timeout: 5m

UEFI / HVF (Apple Silicon and aarch64)

• accel: on Darwin, machine type uses HVF (virt,accel=hvf / q35,accel=hvf). On Linux, KVM with TCG fallback.
• UEFI: aarch64 guests need EDK2 pflash. grain looks for code/vars under Homebrew and common distro paths, e.g.:
  • /opt/homebrew/share/qemu/edk2-aarch64-code.fd
  • /usr/local/share/qemu/edk2-aarch64-code.fd
  • /usr/share/AAVMF/… (Linux)

If firmware is missing, install/reinstall QEMU (brew install qemu) and check QEMU logs. Vars are copied per-VM under the VM dir as flash-vars.fd.

cloud-init

• Seed ISO is seed.iso in the VM dir (NoCloud, volume label cidata).
• macOS builds the seed with hdiutil; doctor checks for it.
• Base userdata sets hostname, SSH keys, ubuntu/grain users, and a grain-ready marker.
• --userdata-file: plain shell is appended as runcmd; #cloud-config merges keys (packages, runcmd, write_files, users append).
• -v mounts add mount -t 9p … runcmds on first seed generation.

If the guest never accepts SSH, serial often shows cloud-init failures (bad YAML in userdata, package errors). Fix userdata and recreate the VM (seed is first-boot oriented).

Image pull / SHA mismatch

sha256 mismatch: got … want …

The partial download is deleted. Causes: interrupted/corrupt download, or Ubuntu moved the release file while the catalog digest is stale. Retry pull; if it keeps failing, update grain (catalog digests) or check network/proxy MITM.

image "…" not pulled (run: grain image pull …)

Run grain image pull before grain new (or let doctor tell you).

Resource caps

Defaults (and overrides in ~/.grain/config.yaml) limit concurrent load. Stopped VMs do not count toward running totals. 0 means unlimited for that knob.

max_vms: 8
max_cpus_total: 16
max_memory_mb_total: 32768
max_cpus_per_vm: 8
max_memory_mb_per_vm: 16384

Errors look like:

resource cap: max_vms is 8 (already 8 running)
resource cap: max_cpus_per_vm is 8 (requested 16)

Stop or remove VMs, lower -c/-m, or raise caps in config and restart the daemon if needed.

stop / start / rm behavior

grain start requires the disk path still present and host mount dirs still valid.

Daemon not up

daemon not up — run: grain up

grain up
grain doctor   # should show socket OK

Port publish rejected

host port 80 is privileged (< 1024)

Use -P 8080:80 or -P 80 (auto high host port). See Networking.

Related

• Networking
• Mounts
• Images