Since this container is used during office hours, I have been testing different methods with test container (similar in size) with no issues whatsoever, but this one in particular keeps failing on the export step. I have also tried to publish image from snapshot and export image using fingerprint generated after publishing image instead of the image name, but had same error. Maybe it’s worth mentioning that the destination directory is an nfs share, however I have also tested export on local directory and error persists.
It would be nice to make this procedure work as it is simple and will have a tar image available to fire up on any other LXD server at any time. I also prefer to backup tarballs on remote nfs share rather than LXD remote as I find it much more flexible.
Since I couldn’t replicate this issue on any other container I am sort of stuck. Do any of you have any idea where to look and why it gets different fingerprint from the one expected?
Also, is there a reason you’re not just doing lxc export <instance> /backups/directory/instance.tar.gz to backup the instance (as opposed to creating an image and exporting that)?
Would probably be good to check what the sha256sum is for the image in /var/lib/lxd/images/ (or /var/snap/lxd/common/lxd/images if using the snap).
That sha256 should line up with what the database has in lxc image list, if it doesn’t, then something got corrupted during publish. If that’s correct, then the issue is the image getting somehow mangled during the download process.
I tried to export from snapshot and also using --force option while publishing, theoretically it shouldn’t be changing unless i’m missing something. What else could cause the image to change during publish/export?