[feature] sync or push should continue after error

[clemenko]

Is this RFE related to an Existing Problem? If so, please describe:

Customer is asking for the hauler sync or push functions continue after an error. Specifically from image pull that didn't work they want hauler to try and complete the rest of the list.

Describe Proposed Solution(s):

Have a flag that will try every item in the manifest.

Describe Possible Alternatives:

Retrying the sync/push until the one item completes.

Additional Context:

[zackbradys]

Similar issue to Issue #284!

[ethanchowell]

It's not so much the image pull or retries that are causing issues as it is the behavior when you try to hauler store copy to a registry with tag immutability configured. In those cases if one of N number of images exists in the target registry, hauler exits at that first failure leaving some number of images unable to push.

[zackbradys]

Hey @ethanchowell... thanks for the additional information. We were discussing this a bit today and will continue to investigate how to address it. ECR is always tricky to work with...

[zackbradys]

Hey @clemenko / @ethanchowell, we resolved one of the issues of continuing on error and maintaining the amount of retries, defaulted at three retries, with PR #317 and v1.0.8, but will keep this open as an Issue/RFE due to the fact we still need to address tag immutability.

[ethanchowell]

I just tested 1.0.8 and I'm still observing the issue where when I run hauler store copy registry://<dest>, if one image fails, execution stops and no other images push. This isn't really related to ecr and it's mechanics around tag mutability, I only used that as an example because it's an easy one to validate. The bigger issue is that if I rely on hauler to an airgapped environment and try to upload those images, I'm dead in the water if one of those image uploads fails because the registry I'm pushing to doesn't allow overwriting images they already have. Looks like it's a result of the bulk write failing here

[zackbradys]

partially similar to #323 and linked others...