<html>
<head>
<meta content="text/html; charset=windows-1252"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 03/12/2015 10:02 AM, Simone Tiraboschi wrote:<br>
<blockquote
cite="mid:1957349372.20499494.1426168924455.JavaMail.zimbra@redhat.com"
type="cite">
<pre wrap="">Hi all,
cloud-init is a powerful tool to configure from outside a cloud instance or an appliance as in our scenario.
Deploying the engine as an appliance is indeed a good way to speed up and make easier the hosted-engine deployment: you don't need to install an OS on the engine virtual machine and than install the engine and so on but you could simply run a ready to use oVirt engine appliance. But you still need to configure it and so cloud-init support within hosted-engine is a reasonable way to complement it.
Then we could also integrate it with oVirt node to let the user input the required info from node TUI in order to have an almost unattended hosted-engine setup on oVirt node using an engine appliance with cloud-init.
The idea is to collect the required information interactively from hosted-engine setup or from node TUI (passing them to hosted-engine setup via an answer file) and pass them to the appliance via cloud-init using a no-cloud datasource.
So now the question is what do you really want to configure via cloud-init?
It's just to define what we want in order to be more focused on user needs:
for instance we could configure engine VM instance hostname, we could set the root password, we could create other users, we could upload ssh private keys, we could run a command on the first boot and so on.
So, if you have any ideas or requirement about that it's the right time for it.</pre>
</blockquote>
<br>
Great suggestions, Simone!<br>
<br>
The things you list cover most of what I do to the engine VM (I do
all of those except for "configure ssh private keys"). In addition,
I:<br>
<br>
<ul>
<li>Add a couple of packages (e.g. zsh)</li>
<li>Configure alternate shells for some initial users (this could
be "run a command on the first boot", as long as adding the
necessary packages was done before that somehow, perhaps simply
as a prior command)<br>
</li>
<li>Populate some specific files, e.g.:<br>
</li>
<ul>
<li>Add specific ssh *public* keys into the
$HOME/.ssh/authorized_keys files for some user and root
accounts</li>
<li>Replace /etc/hosts with a master copy that contains all
hosts on my network<br>
</li>
</ul>
</ul>
<br>
I'd like to see the "additional packages" abstracted out somehow and
added prior to the "run a command on the first boot" step, as
opposed to using "yum" explicitly for one of those commands, but
that's somewhat of a cosmetic "nice to have" since obviously it can
be done explicitly. It's separate conceptually, so would be nice to
treat as such.<br>
<br>
I'm learning Puppet at the moment, and it strikes me that what you
want to do is pretty much the same thing that Puppet manifests are
designed to do.<br>
<br>
Thanks,<br>
Bob<br>
<br>
<blockquote
cite="mid:1957349372.20499494.1426168924455.JavaMail.zimbra@redhat.com"
type="cite">
<pre wrap="">
thanks,
Simone
_______________________________________________
Devel mailing list
<a class="moz-txt-link-abbreviated" href="mailto:Devel@ovirt.org">Devel@ovirt.org</a>
<a class="moz-txt-link-freetext" href="http://lists.ovirt.org/mailman/listinfo/devel">http://lists.ovirt.org/mailman/listinfo/devel</a>
</pre>
</blockquote>
<br>
</body>
</html>