You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
SOCI doesn't use any sort of credential rotation system, so in systems such as Kubernetes, we can encounter issues where a user can attempt to run tasks after the credential has expired. stargz gets around this by only calling the registry if the fetched size is under the expected layer size: containerd/stargz-snapshotter#1584. We can do the same here.
Ultimately, this is a workaround, and we need a better credentials mechanism, but this should be a simple fix that should help for now.
Describe the solution you'd like
No response
Describe any alternative solutions/features you've considered
No response
Any additional context or information about the feature request
There's a separate issue here which is that we don't know exactly when the background fetcher has concluded fetching all of its data. I think if we can implement that, this fix can be a bit cleaner.
The text was updated successfully, but these errors were encountered:
sondavidb
changed the title
[FEATURE] Don't make registry calls when content is fetched
[FEATURE] Don't make registry calls when content has completed fetching
Mar 15, 2024
Does this mean we are not checking the cache before going to the registry? If so, what is the point of the cache?
AFAIK, before this change, the cache still did its function, but the call order meant that we still made a registry call regardless. The cache is to stop us from fetching data we don't already have, and it seems to do its job
Description
SOCI doesn't use any sort of credential rotation system, so in systems such as Kubernetes, we can encounter issues where a user can attempt to run tasks after the credential has expired. stargz gets around this by only calling the registry if the fetched size is under the expected layer size: containerd/stargz-snapshotter#1584. We can do the same here.
Ultimately, this is a workaround, and we need a better credentials mechanism, but this should be a simple fix that should help for now.
Describe the solution you'd like
No response
Describe any alternative solutions/features you've considered
No response
Any additional context or information about the feature request
There's a separate issue here which is that we don't know exactly when the background fetcher has concluded fetching all of its data. I think if we can implement that, this fix can be a bit cleaner.
The text was updated successfully, but these errors were encountered: