I spent a week (or two) using what I could find on the web to write bash scripts to build EBS and local instance store AMIs (it would also automatically launch them so I could immediately test). I have some nice, simple, fast scripts that can build CentOS5 through rawhide that I completely understand (with a good idea of how to convert to my PV images to HVM). But having done that, I find boxgrinder to be a bit more powerful (if slower at times) and, possibly, better solution (I hate reinventing wheels even thought I like to make at least one wheel on my own so I know how one is made.
Having done that and having discovered boxgrinder, I set upon bulding a simple AMI. Took me a few hours to realize that I really need to do a --force after not having things quite right. But in far less time than it took me to write my bash scripts, boxgrinder grinded out a local instance store AMI for me and uploaded to S3.
Then I found there was already the EBS option (cool... since my script started with an EBS and then did the S3 bit).
I diligiently copied the same info that had worked in the S3 section (plugins:->s3:) into the EBS section (plugins:->ebs:) in my config and the EBS build failed gloriously much like the trace listed in the Troubleshooting section of this site: https://twiki.cern.ch/twiki/bin/view/Main/ScalrEDAC (in the section under "ERROR with Amazon account number").
Once I hyphenated my Amazon account number, life was good again.
So, my questions are:
-- when it comes to Amazon account number (account_number:DATA), is the recommended practice to use the hypenated or non-hypenated version?
-- is there a way we can combine the S3/EBS data into a single (AMI?) section so it doesn't have to be repeated?
- is there an HVM delivery option yet that I haven't found?
So far I'm loving this... I have a simple development platform running mock and boxgrinder now and see lots of potenIial. I'm just trying to get an feel for what I need to workaround and/or do myself.
Thanks.
