On 7/15/21 8:01 PM, Do Not Reply wrote: > Detected one or more Cirrus-CI cron-triggered jobs have failed recently: > > Cron build '1.21' Failed: https://cirrus-ci.com/build/6167949355515904 Ed, et al., Do you know this test? I don't understand how this could be a flake, it seems like a bug...but it's on a release branch: https://cirrus-ci.com/task/5576282244120576?logs=integration_test#L944 Shouldn't it be trying to pull 'alpine' (all the messages refer to 'base')?

On 7/16/21 3:29 PM, Ed Santiago wrote: > > Shouldn't it be trying to pull 'alpine' (all the messages refer to 'base')? > > The error I see is: > > Error initializing source docker://registry.fedoraproject.org/base:latest > > Any time I see 'registry.fedoraproject.org', my first reaction > is: flake. It's an unreliable registry. I can't check right now > if that particular test uses _prefetch; and in fact I don't > really know if _prefetch is still working as intended. I will > look into it next week. Ya, I saw that error at first also, I think it's wrong and misleading. That image doesn't exist: ~ $ skopeo inspect docker://registry.fedoraproject.org/base:latest FATA[0000] Error parsing image name "docker://registry.fedoraproject.org/base:latest": Error reading manifest latest in registry.fedoraproject.org/base: manifest unknown: manifest unknown If you look a bit further up, you'll see it's trying build-time alias `alpine`: STEP 1: FROM alpine AS base By my reading, we should have pulled this from quay.io/libpod (b/c registries.conf). Maybe quay was broken at the time (it does happen). Still, I think there's a secondary issue here as well. Even if the registries were being wonky, not able to pull down `alpine`, IMHO it's completely broken that the error reports 'base' as the failed image. If this has confused two adept users, newbies would stand zero chance of decoding this. I can open an issue, assuming you concur.