Trackhub S3 Integration

[gpratt]

Hi,

In the course of my work we've had more than a number of trackhub servers go down. Our lab is thinking about migrating out trackhubs over to amazon S3 hosting, so I wrote a quick afternoon hack up to see if it is possible. Long story short it is.

This code very much isn't ready to be integrated into the main trackhub repo, but I was wondering, would other people be interested in this functionality, and would it be useful to fix up the code to make it more general? More importantly what should the interaction between S3 and trackhub look like. Login credentials to S3 are a complex thing.

Thanks,

Gabe

---

This change is

[daler]

Good idea. I probably wouldn't use this myself, but it should be easy to add and might make the package more useful to others.

I think the interaction between S3 and trackhub should be functionally similar to tools like cloudbiolinux, starcluster, s3cmd, aws-cli, etc. That is, check for credentials in env vars, or if not found read them from a particular config file.

http://docs.aws.amazon.com/cli/latest/userguide/cli-chap-getting-started.html lays out the details of where aws-cli looks for credentials; I think this strategy would work for trackhub.