[chef] Re: Re: Re: Re: Re: Re: Re: Re: Re: Scaling erchef horizontally

[Victor Iglesias < >]

This is great! 

I am going to try this against my Eucalyptus install as well. 

On Fri, Apr 25, 2014 at 9:29 AM, Julian C. Dunn < " target="_blank"> > wrote:

We haven't documented it because it's not officially supported, but
basically the directions are:

1. Make a bucket and an IAM user
2. Grant "s3:PutObject", "s3:GetObject", "s3:DeleteObject", and
"s3:ListBucket" for said bucket to the IAM user
3. Modify config settings in /etc/opscode/private-chef.rb (or the
corresponding variant for OSS Chef Server)
  a) bookshelf['vip'] = 's3.amazonaws.com' # or whatever
region-specific endpoint you want
  b) bookshelf['access_key_id'] = 'your IAM user access key'
  c) bookshelf['secret_access_key'] = 'your IAM user secret key'
  d) opscode_erchef['s3_bucket'] = 'bucket_name'
4. private-chef-ctl reconfigure
5. optionally disable bookshelf locally as you won't need it.

regards,
Julian

On Fri, Apr 25, 2014 at 11:59 AM, Ranjib Dey < "> > wrote:
> julian,
> is there any documentation on this, i.e. how to use s3 as cookbook store?
> what are the config changes required etc?
> regards
> ranjib
>
>
> On Fri, Apr 25, 2014 at 7:05 AM, Julian C. Dunn < "> > wrote:
>>
>> BTW, in case it wasn't mentioned before: Since Bookshelf speaks the S3
>> protocol, it is possible to have erchef use S3 directly and turn off
>> Bookshelf entirely.
>>
>> - Julian
>>
>> On Fri, Apr 25, 2014 at 2:59 AM, Cameron Cope < "> >
>> wrote:
>> > I am currently running a Chef 10 cluster in EC2 with the bookshelf
>> > stored on
>> > a Gluster volume mounted on all the servers, but only one node at a time
>> > is
>> > receiving traffic for rabbitmq, solr, or couchdb. I've been holding off
>> > on
>> > upgrading to Chef 11 server because I haven't had time to figure out HA,
>> > so
>> > I would be very interested in seeing your cluster configuration for Chef
>> > 11.
>> >
>> > From what I understand, it is possible to run rabbitmq in an HA mode
>> > where
>> > the queues are mirrored to slaves, but during failure there can be some
>> > consistency loss.
>> > http://www.rabbitmq.com/ha.html
>> >
>> > I think solr can be made HA as well, but it looks complicated:
>> > https://cwiki.apache.org/confluence/display/solr/SolrCloud
>> >
>> > When I last checked into pgpool, I got scared off by the SQL
>> > restrictions,
>> > but if it works and is stable that would be awesome!
>> > http://www.pgpool.net/docs/latest/pgpool-en.html#restriction
>>
>> --
>> [ Julian C. Dunn < "> >          * Sorry, I'm    ]
>> [ WWW: http://www.aquezada.com/staff/julian    * only Web 1.0  ]
>> [ gopher://sdf.org/1/users/keymaker/           * compliant!    ]
>> [ PGP: 91B3 7A9D 683C 7C16 715F 442C 6065 D533 FDC2 05B9       ]
>
>



--
[ Julian C. Dunn < "> >          * Sorry, I'm    ]
[ WWW: http://www.aquezada.com/staff/julian    * only Web 1.0  ]
[ gopher://sdf.org/1/users/keymaker/           * compliant!    ]
[ PGP: 91B3 7A9D 683C 7C16 715F 442C 6065 D533 FDC2 05B9       ]