Brad Durrow

@fdr, Thank you for your feedback. When you suggest writing a prototype first what do you mean?

I think that reworking the retry mechanism to be more robust in dealing with transient failures would help a lot. Something like this is what I am thinking of: https://cloud.google.com/storage/docs/exponential-backoff...

I don't think that my workaround is a good solution but it gets my backups into GCS. I modified blobstore/gs/utils.py so that blob.upload_from_file is called with num_retries=10 instead of num_retries=0....

It looks like 0.18.3 is in my virtualenv. By reading the code I don't know how but my expectation is that the the [way that wal-e calls the client library...

When the backup-push fails due to this condition it does not show up in backup-list.

We use S3 as object storage and Postgres 9.4. Even with 9.4 we don't get enough data from the basebackup to cleanly start; we receive an error message suggesting we...

What do you think about reusing the existing logic for the webserver that retrieves the file to the local filesystem then if a header is set on the request offers...

How do you imagine the unit tests working? There are two "remote" relationships, in our use case they are represented by STAGE -> BACKUP and BACKUP -> S3. I haven't...

> The header thing you're proposing for selecting which way to save the file sounds ok to me though I'm not sure what you meant by: "After pghoard_postgres_command successfully retrieved...

I believe I need to create test_storage_configs.py to allow for object store testing. Can you provide an example for how to set this up?