First time integration sanity check!

I’m about to start testing my image on the demo server before moving to AWS and I’m a bit confused with the documentation. Hope no one minds what are probably stupid questions!

This is what I am planning on doing:-

  1. Create my golden image for Pi CM4 headerless system. V1.00
  2. Use mender-convert to create the img and artifact.
  3. Write this mender img to a virgin card and deploy on CM4.
  4. Upload the artifact to the server - not sure if this is necessary?
  5. Make a change to the golden-image - just change the led colour so I can see it has happened.V1.01
  6. Use mender-convert to create the new img and artifact.
  7. Upload the 1.01 artifact to the server.
  8. Deploy to test hardware.

Does this all look sensible?
Couple more questions.

  1. What happens if a device is switched off when 1.01 and 1.02 are deployed and is only switched on when 1.02 is ready? Do the intermediate artifacts have to be deployed in order?
  2. Any ideas what server on AWS I will need? We are producing 10,000 copies of the hardware for the first production run.

Thanks for any help,


Hi @jjsteele,

The flow as described is correct.

For your questions:

  1. that depends, if you have declared the Artifacts having dependencies to each other or not. Taking the simplest case (no dependencies, full rootfs upgrade), 1.02 can be deployed to a device running 1.00, 1.01, whatever. Please note that the key is not if the device is powered on or off - it is which deployments are active when the device is connected. See the documentation for more details.
  2. we are happy to provide suitable backend service for device fleets of that size, optionally also including SLAs. Please get in touch with us for that.
    2.1. if you want to go self-hosting, well, it depends. :wink: