[chef] Re: Re: Re: Re: managing chef server/webui/couchdb with chef?

[Michael Callahan < >]

OK, being Chef novices, we weren't clear whether there was anything particularly clever needed to upgrade and restart a running chef server.  Perhaps not, since I suppose the restart would be initiated in any case by a chef-client or chef-solo process which is basically independent of the chef-server it's restarting (though it's also not clear to me whether or not the chef-client or chef-solo might not subsequently try to make requests to the chef-server process e.g. for data-bags used by later resources while it is rebooting, and what would happen if these requests failed with an ECONNREFUSED).

Also, assuming that you're referring to the chef/recipes/server.rb recipe, we've actually had difficulties with it failing to restart chef-server at all if you are running chef-server-webui due, I think, to COOK-364.

Thanks,

Michael

On Mon, Oct 11, 2010 at 2:00 AM, Jon Wood < "> > wrote:

The Opscode repository contains a cookbook to manage Chef, which has a
server recipe included - the same one used when doing a chef-solo
bootstrap of the server.

That's capable of keeping your Chef server up to date, and runs
maintenance tasks to reduce the size of your CouchDB database.

Jon

On 10 October 2010 21:16, Jim Hopp < "> > wrote:
> Right, we do use chef-solo for the initial install of chef server. I'm
> wondering how people manage on-going updates to the chef server components.
> But I guess we won't be changing the chef server components very often; most
> of the changes to the machine will be the standard machine config stuff.
> Thanks for the affirmation that we can just use chef-client for that.
>
> -Jim
>
> On 10/10/2010 12:35 PM, "> wrote:
>
> Knife has a bootstrap option to setup a box as a chef client which you can
> use chef solo to then run the bootstrap cookbook for server components.
>
> There's no magic with chef server and client living on the same machine
> except that the validation key is already on the machine so it saves you a
> step
>
> On Oct 10, 2010 11:49 AM, "Jim Hopp" < "> > wrote:
>> Hi folks-
>>
>> We're curious about how people are managing the chef server components
>> (Chef Server API/webui, CouchDB, Solr Indexer, rabbitMQ, etc) and the
>> machine all this is running on.
>>
>> First, the machine: it seems feasible to assign a role to the node
>> running the server, and to use that role to manage the prosaic stuff:
>> packages, iptables, ntp, accounts, etc; all the stuff we're managing on
>> the all the other nodes. Are there any gotcha's to just running
>> chef-client on the chef server node?
>>
>> Second, the chef server components: this seems much more complicated. Do
>> people try to manage updates to the chef server components with chef
>> itself, or do you simply use chef-solo to do the initial install and
>> then manage updates by hand?
>>
>> -Jim
>

--
Jon Wood
Blank Pad Development

07827 888143