David Steele

Info should be correct. This has to do with the manifests stored in `backup.history`. We expect backup labels to be ascending (even vs history), but the introduction of adhoc expire...

> having repo-retention-history=0 to retain the backup history only for unexpired backups could be a valid workaround for us Ah, yes, that should work for now.

I think we need two things here. First, there should be a check in the storageNew() to make sure the interface function has been implemented when links are supported, something...

Ideally we would also implement this method in the remote storage driver. At the moment we don't support creating links remotely, but seems better to write the entire feature rather...

> Having a storageFeature query looks like a better way to go. `storageFeature()` already exists and it is the driver that determines what features are available. Unless I'm misunderstanding?

> As a corner case for the Posix driver, do we need to worry about remote SMB mounts? We have the `cifs` driver for this case, which is basically the...

> I noticed the remote driver doesn't implement .followLinks Not sure what you mean. `followLinks` is in `storageRemoteInfo()` and it does appear to be implemented. > It would be nice...

The next release adds the repo-storage-upload-chunk-size option (https://github.com/pgbackrest/pgbackrest/commit/b05d31f531b55b747175cf94c963f95574c54d2b) which can be configured up to 1TiB. Setting this value too high can cause failures due to lack of memory and limit...

> But in practice setting the chunk size to 1TiB will disable multi-part upload. Although now that I think about it this is not practical since we pre-allocate the buffer....

Thanks for the update. I do think this would make the extension more reliable.