Labgrid useful commands¶

Copy-paste labgrid-client usage for the labgrid-fcefyn coordinator. Place names and exporter layout live in libremesh-tests (ansible/files/exporter/labgrid-fcefyn/exporter.yaml). Ansible deployment: Ansible and Labgrid. SSH to DUTs: SSH access to DUTs.

Runtime context¶

Context	`LG_PROXY`	Typical use
On the lab host (or SSH session to it)	Unset	`who`, `places`, manual `unlock` (talks to local coordinator).
CI runner or machine via SSH jump	`ssh://labgrid-fcefyn` (or `labs.<lab>.proxy` from openwrt-tests labnet.yaml)	`reserve`, pytest, workflows that expect the remote proxy.

For who / places / freeing stuck places, run on the host without LG_PROXY so the client hits the coordinator directly.

Coordinator inspection¶

Command	Purpose
`labgrid-client who`	Which user holds which place (acquired / reserved).
`labgrid-client -v places`	Verbose place list; shows acquired and owner (`labgrid-fcefyn/user`).
`labgrid-client places`	Short place list.

Places (labgrid-fcefyn)¶

Top-level keys in exporter.yaml (use as -p argument):

Place name
`labgrid-fcefyn-belkin_rt3200_1`
`labgrid-fcefyn-belkin_rt3200_2`
`labgrid-fcefyn-belkin_rt3200_3`
`labgrid-fcefyn-bananapi_bpi-r4`
`labgrid-fcefyn-openwrt_one`
`labgrid-fcefyn-librerouter_1`

Lock and unlock a place¶

Same verbs as libremesh-tests workflows (lock after reserve token, unlock in always).

labgrid-client -p labgrid-fcefyn-bananapi_bpi-r4 lock
labgrid-client -p labgrid-fcefyn-bananapi_bpi-r4 unlock

Unlock every stuck place after a crashed session or aborted manual test:

for p in \
  labgrid-fcefyn-belkin_rt3200_1 \
  labgrid-fcefyn-belkin_rt3200_2 \
  labgrid-fcefyn-belkin_rt3200_3 \
  labgrid-fcefyn-bananapi_bpi-r4 \
  labgrid-fcefyn-openwrt_one \
  labgrid-fcefyn-librerouter_1
do
  labgrid-client -p "$p" unlock
done

Symptom: reserve stays in waiting / matches nothing in CI. Cause: all matching places are locked by someone. Coordinator only assigns free places. Fix: unlock as above (or identify the holder with who), then re-run the job.

Release between test runs (host or dev machine)¶

After a crashed pytest, an aborted mesh run, or before the next manual/CI run, places can stay acquired or locked by the previous session. Run release (same as unlock) on the lab host (coordinator local, LG_PROXY unset; see Runtime context) or from a developer machine with LG_PROXY set if that machine has coordinator access to the same lab.

-k / --kick: releases the place even if another user currently holds it. Use when the holder is stale (dead SSH, killed pytest). On a shared lab, coordinate before kicking someone else's session.

Examples for two common mesh nodes:

labgrid-client -p labgrid-fcefyn-openwrt_one release -k
labgrid-client -p labgrid-fcefyn-librerouter_1 release -k

To clear every listed place in one go (same pattern as the unlock loop above):

for p in \
  labgrid-fcefyn-belkin_rt3200_1 \
  labgrid-fcefyn-belkin_rt3200_2 \
  labgrid-fcefyn-belkin_rt3200_3 \
  labgrid-fcefyn-bananapi_bpi-r4 \
  labgrid-fcefyn-openwrt_one \
  labgrid-fcefyn-librerouter_1
do
  labgrid-client -p "$p" release -k
done

Reservations (CI and manual)¶

Typical split used in workflows (some forks return the reservation token before allocation is final):

labgrid-client reserve --shell '<filter>'   # e.g. device=bananapi_bpi-r4
labgrid-client wait '<token>'

Adjust filters to match places/resources in the suite.

Stuck GitHub Actions `labgrid-client wait`¶

If canceling the workflow does not stop the job, on the runner host:

pkill -9 -f "labgrid-client wait"

Then unlock any orphaned places from the lab host (see loop above).

Host services look broken after reboot¶

If SSH to DUTs times out while ping works, coordinator/exporter/config may be stale. Redeploy Labgrid stack on the host:

ansible-playbook playbook_labgrid.yml -l labgrid-fcefyn

(Playbook path and inventory as in ansible-labgrid.)

VLAN or switch issues are outside labgrid-client; see Routine operations - DUTs and VLANs and switch-config.