Hi Adrian, This is a known problem, and I'd like to remove the default dependency on S3 for hosting scripts. There's a proposal on https://issues.apache.org/jira/browse/WHIRR-225 for how to migrate away from this. Please let us know your thoughts!
In the meantime, you can set the whirr.run-url-base property to point to another S3 bucket (or any webserver) to host a variant of the scripts for testing/customization. Cheers, Tom On Sun, Feb 6, 2011 at 12:13 PM, Adrian Cole <[email protected]> wrote: > Hi, guys. > > I'm sure that there's been discussion about this, but I wanted to call > attention to something. Implementing WHIRR-124 was pretty tough due to > needing changes to runurl scripts. In WHIRR-124, provider keys are now > different, and these are hard-coded in the runurl scripts. I don't even > know where the version control is for these scripts. I'd like to be able > to change my local copy of hadoop's post-configure and not have to worry > about who's using the public s3 bucket at the time, and I'd like to push > changes to runurl scripts as a part of a patch. > > How close are we to being able to isolate tests to only use files or code in > a local checkout of svn? > > Cheers, > -Adrian >
