The problem with scripts pushed as userdata is that as far as I know we have
no mechanisms for tracking the execution & collecting the output. jclouds
provides
such tools only for regular scripts executed over ssh.

On Mon, Mar 5, 2012 at 9:04 PM, Ashish <[email protected]> wrote:

> I need to some custom operations on AMI like privileges, mounting
> drives with specific names, used by my scripts.
>
> Don't want to mix them with existing whirr functions, as it will make
> it hard to maintain and migrate them.
>
> To add a new function would need code change, as script names are
> hardcoded inside HadoopClusterActionHandler
>
> Bootstrap would be a cleaner way to do this, rather than using user-data.
>
> Is there any limitation that i should be aware of. Using bootstrap
> way, I shall be able to execute script on instances, right before
> Hadoop goes live :)
>
> On Tue, Mar 6, 2012 at 12:27 AM, Andrei Savu <[email protected]>
> wrote:
> >
> >
> > On Mon, Mar 5, 2012 at 7:40 PM, Ashish <[email protected]> wrote:
> >>
> >> They can essentially help you do some stuff on the instance like
> >> attaching EBS etc.
> >> I am not sure if beforeBootstrap() function to achieve this.
> >
> >
> > Yes! You should be able to do this by customising the bootstrap
> > (install_*.sh) script for a service.
>
>
>
> --
> thanks
> ashish
>
> Blog: http://www.ashishpaliwal.com/blog
> My Photo Galleries: http://www.pbase.com/ashishpaliwal
>

Reply via email to