I should have made this clearer: the idea it let whirr manage ssh_config
sensibly: only the sections between #### BEGIN cluster-name #### and ####
END cluster-name ####
will be changed by whirr. If those lines don't exist, they are added at the
end of the ssh config file.
The actual ssh config file could be controlled by a property. I don't have
a strong opinion on the default value: ~/.whirr/{cluster}/ssh-config or
~/.ssh/config...
Regards,
Karel
On Tue, Mar 27, 2012 at 11:19 AM, Andrei Savu <savu.and...@gmail.com> wrote:
> I like this but we need to put the generated file under ~/.whirr/
>
> I already have a customised version of ~/.ssh/config I wouldn't want to
> mess with that.
>
> ssh -F ~/.whirr/<cluster-name>/ssh-config <public-ip-hostname>
>
> How about generating entries with the role names? It would be great if we
> could do something like this:
>
> ssh -F ~/.whirr/hadoop/ssh-config hadoop-namenode
>
> -- Andrei Savu
>
> On Tue, Mar 27, 2012 at 11:30 AM, Karel Vervaeke <ka...@ngdata.com> wrote:
>
> > Hey,
> >
> > how would you feel about whirr generating a bit of ssh configuration for
> > you?
> > This would make it simpler to log into your whirr-launched machines
> (simply
> > 'ssh publichostname')
> > Tools like http://code.google.com/p/parallel-ssh/ also depend on a
> > correctly setup passwordless ssh, so those would benefit as well.
> >
> > The idea would be to inject something like this:
> >
> > ###### BEGIN WHIRR your-cluster-name ######
> > # This is automatically updated by whirr. Your puny manual editing hands
> > stand no chance
> > Host ec2-12.23.45.67.ec2.amazon-aws.com
> > User yourclusteruser
> > IdentityFile yourwhirr-private-key
> > PasswordAuthentication no
> > UserKnownHostsFile ~/.whirr/your-cluster-name/known_hosts
> >
> > I'm not sure yet about the UserKnownHostsFile. We could just use
> /dev/null
> > + StrictHostKeyChecking = no, but that's less secure.
> >
> > WDYT?
> >
>
