omf loadLoad a pre-built disk image onto one or more COSMOS testbed nodes so they boot
into a known software environment ready for your experiment.
Before running any experiment on COSMOS you must install a disk image onto the
nodes you reserved. The omf load command automates this: it PXE-boots the
target nodes, streams the compressed image from the testbed image repository,
writes it to the node's local disk, and powers the node off when done.
Every COSMOS tutorial begins with an omf load step. This page explains the
command in isolation so that newcomers understand what it does before following
a domain-specific tutorial.
After completing this tutorial you will be able to:
omf load to install a named image onto one or more testbed nodes.omf tell and omf stat to power nodes on and verify they are up after| Difficulty | Beginner |
| Estimated time | 15–30 min (image transfer time varies by image size) |
| Domain / sandbox | Any COSMOS domain with a reservation |
| Topic group | Getting started |
| Last verified | Not re-tested (migrated 2026-06-20) |
Background knowledge
omf load is step 3 in that workflow.Account & access
Devices / nodes
| Resource | Role | Qty | Notes |
|---|---|---|---|
| TODO: verify | Target node(s) to image | 1+ | Any reserved node in the domain |
Disk images
| Image | Load onto | Provides |
|---|---|---|
| TODO: verify | All target nodes | The specific software environment needed for your experiment (e.g. baseline.ndz provides a clean Ubuntu base) |
Software components
| Component | Version | Source |
|---|---|---|
omf-expctl (omf CLI) |
TODO: verify | Pre-installed on all COSMOS consoles |
Not applicable. omf load operates on the testbed control plane: the console
contacts the PXE/imaging service, which streams the image to the target nodes
over the management network. No data-plane or RF topology is involved at this
stage.
Reserve the nodes (see Prerequisites) and log into the domain console:
ssh <username>@console.<sandbox>.cosmos-lab.org
Power the nodes off before imaging (prevents in-flight disk writes):
omf tell -a offh -t system:topo:allres
Run omf load with the image name and a node target. To image all reserved
nodes at once:
omf load -i baseline.ndz -t all
To image a single node:
omf load -i baseline.ndz -t node1-1
The command PXE-boots each target node, streams the compressed image, writes it
to the local disk, and powers the node off automatically when imaging is
complete.
Note: Replace
baseline.ndzwith the image required for your experiment.
See the image registry for available
images, or ask your testbed administrator.
After omf load finishes, turn the nodes back on:
omf tell -a on -t all
Check that all nodes have booted and are reachable:
omf stat -t all
Wait until all nodes report a ready/up status before proceeding with your
experiment. It typically takes 1–3 minutes for nodes to complete boot after
imaging.
From the console, SSH to a node to confirm the image is correct:
ssh root@node1-1
omf load completes without errors for all target nodes.omf stat -t all shows all nodes as up / ready./etc/omf-image-name if your image stamps that file, or run aIf you did not modify the node's disk during your experiment, simply power
the nodes off at the end of your reservation:
omf tell -a offh -t system:topo:allres
If you installed packages or generated data you want to keep, save the image
before your reservation ends:
omf save -n node1-1
See the OMF commands reference for full
omf save options.
| Symptom | Likely cause | Fix |
|---|---|---|
omf load hangs or times out |
Node did not PXE-boot (legacy BIOS boot order may pick local disk first) | Power-cycle the node with omf tell -a reset -t <node>, or contact testbed support to set the PXE boot flag |
Node shows as failed in omf stat |
Image transfer error or disk write failure | Re-run omf load for the affected node; if it persists, try a different image |
Node does not appear in omf stat |
Node did not come up after omf tell -a on |
Wait 2–3 more minutes; if still missing, power-cycle with omf tell -a reset -t <node> |
| Wrong image boots | Another disk on the node still has a bootable OS | Contact testbed support; multi-disk boot-order conflicts require manual intervention |
omf load, omf tell, omf stat, omf save)Author(s): COSMOS team · Last verified: Not re-tested (migrated 2026-06-20) · Tested image/release: TODO: verify · Tags: omf, imaging, getting-started, omf-load