Frequent 502 Bad Gateway and 504 Gateway Timeout errors.

When using oc mirror --v2 without limiting the number of concurrent executions, 500 errors occur with a significant probability when pulling from public registries (quay.io, registry.redhat.io) to local or from local to mirror registries. Is oc mirror v2 behaving poorly?

As explained in 1 and 2, if it fails once, it does not retry and proceeds to the end, exiting with exit(0), so it is necessary to strictly limit the number of simultaneous executions.

Especially, the mirror registry built with Single node Quay seems to be rate-limited, as errors occur frequently, so pushes must be done carefully one image at a time.

4.18

Always

    1.
    2.
    3.
    

The following options have fewer failures. (--retry-times is not working. https://issues.redhat.com/browse/OCPBUGS-54587)

Mirroring from container registry to disk
oc mirror --dry-run -c imageset-config.yaml file://mirror_images --v2
oc mirror --retry-times 10 --retry-delay 30s --image-timeout 20m --parallel-images 4 --parallel-layers 5  -c imageset-config.yaml file://mirror_images --v2

Mirroring from disk to mirror registry
oc mirror --retry-times 10 --retry-delay 60s --image-timeout 30m --parallel-images 1 --parallel-layers 10 -c imageset-config.yaml --from file://mirror_images docker://mirror.$cluster_name.$base_Domain:8443 --v2

Mirroring from container registry to mirror registry
oc mirror --dry-run -c imageset-config.yaml --workspace file://mirror_images docker://mirror.$cluster_name.$base_Domain:8443 --v2
oc mirror --retry-times 10 --retry-delay 30s --image-timeout 20m --parallel-images 1 --parallel-layers 10 -c imageset-config.yaml --workspace file://mirror_images docker://mirror.$cluster_name.$base_Domain:8443 --v2