On Thu, Nov 3, 2011 at 6:43 PM, Paolo Castagna < [email protected]> wrote:
> Maybe I ask a stupid question, apologies in advance (*): is it possible to > create an instance/image with all the software available on it and then > somehow clone/duplicate that instance? > It is possible but you need to heavily customise the install & configure scripts to be able to use it. One quick thing you can try is to create an AMI that already has the Hadoop tarball and use the remote:///path/to/tarball.tar.gz addressing scheme for whirr.hadoop.tarball.url OR use S3 for caching the tarballs from your local machine - Whirr can this for you if you specify a whirr.hadoop.tarball.url that starts with file:// - by using a temporary bucket. Also tweak the jclouds.ec2-ami-query as Paul suggested: jclouds.ec2.ami-query=owner-id=$AWS_ACCOUNT_NUMBER;state= available;image-type=machine;root-device-type=instance- store;architecture=x86_32 With this two changes you reduce the chance of failure due to download errors and also reduce the number of API calls jclouds is making.
