(WIP) Add retry of EVE image update to API

[sadov]

If EVE image update fails due to some fluke (such as a power outage) we can currently retry by doing a eveimage-remove, wait until the partition in EVE is marked as "unused", and then redo the eveimage-update.

That three step is hard for the controller to do.

We can add a retry/version count for the EVE images. One way to do that is to use the Version field in message BaseOSConfig { UUIDandVersion uuidandversion = 1;

and if that version changes we would ignore the fact that the image exists as INPROGRESS in some partition.

[sadov]

Yes, this test just repeat the use case with interrupted update. We need to change EVE and then change test for checking of reupdate.

[eriknordmark]

Yes, this test just repeat the use case with interrupted update. We need to change EVE and then change test for checking of reupdate.

Are you doing a PR against EVE for that, or should we create a separate story for that?

[sadov]

Are you doing a PR against EVE for that, or should we create a separate story for that?

The baseos update machinery in EVE (and especially the suggestion to reuse UUIDandVersion to indicate re-update) is not clear for me. Maybe a separate story would make sense.

[eriknordmark]

See https://github.com/lf-edge/eve/pull/2013

[sadov]

@eriknordmark tests updated accordingly https://github.com/lf-edge/eve/pull/2013:

• tests/update_eve_image/testdata/reupdate_eve_image.txt
• tests/update_eve_image/testdata/reupdate_eve_image_oci.txt

This time I get a config with information about the current baseos update, then I send that config with an empty "base" array, and then I send the first one again.

At this point, the tests have failed because an automatic reboot has not been performed after a configuration change. Logs: reupdate_eve_image.048.log reupdate_eve_image_oci.049.log

Maybe in this case we need to change something else in the EVE config?