cloudstack-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Ian Duffy <...@ianduffy.ie>
Subject Re: [GSOC2014] Acceptance of three projects for CloudStack
Date Fri, 25 Apr 2014 20:48:24 GMT
>
> Can I twist your arm to try to use Chef.
> Mainly because we seem to have more folks using Chef than Puppet and
> because we can get help from Chef to optimize the recipes.
> The idea being that once this is all working in a multi-VM setup, users
> could use the recipes in a production setup.


No problem.


On 25 April 2014 21:09, Sebastien Goasguen <runseb@gmail.com> wrote:

>
> On Apr 25, 2014, at 4:04 PM, Ian Duffy <ian@ianduffy.ie> wrote:
>
> >>
> >> Jon created a xenserver-6.2-sp1 vagrant box (manually) and uploaded it
> to
> >> vagrant cloud:
> >>
> >> https://vagrantcloud.com/jonludlam/xenserver-6.2-sp1
> >>
> >> (It looks like your packerfile should cleanly automate the production of
> >> this — is that right?)
> >>
> >> I created a very simple ‘infrastructure’ VM which contains mysql, NFS
> for
> >> primary and secondary and NAT/DNSMASQ so it can act as the gateway for
> the
> >> nested guests:
> >>
> >>
> >>
> https://github.com/djs55/cloudstack-dev-vms/tree/master/cloudstack-infrastructure
> >>
> >> and I customised Jon’s box to switch to bridge, fiddle with networking
> >> settings etc:
> >>
> >> https://github.com/djs55/cloudstack-dev-vms/tree/master/xenserver-6.2
> >>
> >> and I’ve got a marvin/deployDatacenter config:
> >>
> >> https://github.com/djs55/cloudstack-dev-vms/blob/master/datacenter.cfg
> >
> >
> > Awesome stuff! My packerfile does generate that .box automatically. It
> > gives the NAT interface for vagrant and a host-only network on
> > 192.168.56.10 (matching current devcloud.cfg). 192.168.56.10 traffic is
> > forwarded to the NAT interface to allow the VMs to get internet.
> >
> > I to plan to have a MySql + NFS box. Not sure whether to do this as a
> baked
> > image created with packer or a CentOS base box provisioned on first boot
> by
> > puppet. It would be nice for the management server to run on here to for
> > those who don't wish to run it on their host os. The ram requirements
> seem
> > a bit high though. (2gb for management + mysql + nfs, 4gb for xenserver).
> >
>
> Can I twist your arm to try to use Chef.
> Mainly because we seem to have more folks using Chef than Puppet and
> because we can get help from Chef to optimize the recipes.
> The idea being that once this is all working in a multi-VM setup, users
> could use the recipes in a production setup.
>
> > 3. I don’t know how best to spawn multiple xenservers, manage their IPs,
> >> add them to the cloud etc.
> >
> >
> > Haven't looked into this too much. I believe it should be possible to use
> > the multi-machine feature in vagrant for this:
> >
> > https://docs.vagrantup.com/v2/multi-machine/
> >
> >
> http://russellcardullo.com/blog/2012/09/08/setting-up-a-multi-vm-environment-in-vagrant/
> >
> > Plus of course to test KVM we need a hypervisor supported by vagrant and
> >> which supports nested virt which probably means supporting running also
> on
> >> a KVM host. So far our Vagrantfiles and boxes are virtualbox-specific.
> >
> >
> > Solution would be to use vmware fusion/workstation (host OS dependant) It
> > will allow for nested virtualization, however it will have licensing
> > costs(Vagrant Plugin + Vmware Product).
> >
> >
> > On 25 April 2014 19:07, Dave Scott <Dave.Scott@citrix.com> wrote:
> >
> >> Hi,
> >>
> >> On 25 Apr 2014, at 17:41, Ian Duffy <ian@ianduffy.ie> wrote:
> >>
> >>> Hi All,
> >>>
> >>> Thanks for the congratulations! Looking forward to putting more time
> >>> into Cloudstack.
> >>>
> >>> My project in brief involves re-creating devcloud using vagrant and
> >>> provisioning them with puppet. There's a full proposal at
> >>>
> >>
> http://www.google-melange.com/gsoc/proposal/public/google/gsoc2014/imduffy15/5662278724616192
> >>> for anybody interested.
> >>
> >> I think this will be really good!
> >>
> >>> I've started doing a small amount of work on it. As stated in the
> >>> proposal I'm going to be using xenserver. I read up on doing
> >>> unattended installations with xenserver and created a packerfile for
> >>> creating the box file https://github.com/imduffy15/packer-xenserver
> >>
> >> By coincidence Jon Ludlam (cc:d) and I have started working on something
> >> similar — maybe we can collaborate. Our goal was to be able to create
> >> vagrant definitions (or similar) that we could glue together like lego,
> to
> >> make development environments with (e.g.) multiple xenservers to test
> >> migration and environments with both xen and kvm to test interop, while
> >> being able to run the management server locally for ease of debugging…
> well
> >> that’s the dream anyway :-)
> >>
> >> Jon created a xenserver-6.2-sp1 vagrant box (manually) and uploaded it
> to
> >> vagrant cloud:
> >>
> >> https://vagrantcloud.com/jonludlam/xenserver-6.2-sp1
> >>
> >> (It looks like your packerfile should cleanly automate the production of
> >> this — is that right?)
> >>
> >> I created a very simple ‘infrastructure’ VM which contains mysql, NFS
> for
> >> primary and secondary and NAT/DNSMASQ so it can act as the gateway for
> the
> >> nested guests:
> >>
> >>
> >>
> https://github.com/djs55/cloudstack-dev-vms/tree/master/cloudstack-infrastructure
> >>
> >> and I customised Jon’s box to switch to bridge, fiddle with networking
> >> settings etc:
> >>
> >> https://github.com/djs55/cloudstack-dev-vms/tree/master/xenserver-6.2
> >>
> >> and I’ve got a marvin/deployDatacenter config:
> >>
> >> https://github.com/djs55/cloudstack-dev-vms/blob/master/datacenter.cfg
> >>
> >> Using virtualbox on OSX I can ‘vagrant up’ the infrastructure VM,
> ‘vagrant
> >> up’ the xenserver and then run the management server locally… just
> about.
> >> There are still lots of rough edges like
> >>
> >> 1. I think I need to do something to ensure the management server uses a
> >> sensible IP address; since I’m running it on the host for ease of
> >> development, there are about 10 different IPs it could choose
> >>
> >> 2. When switching between cloudstack branches I need to download a
> >> matching system VM template. I was wondering if I could auto-download
> the
> >> right template directly into NFS secondary storage on the infrastructure
> >> VM. I keep forgetting to do this and it’s very annoying!
> >>
> >> 3. I don’t know how best to spawn multiple xenservers, manage their IPs,
> >> add them to the cloud etc. The deployDatacenter .json format seems to be
> >> being actively developed, so now the one in the repo only works with
> >> cloudstack/master. Maybe this is where puppet comes in?
> >>
> >> Plus of course to test KVM we need a hypervisor supported by vagrant and
> >> which supports nested virt which probably means supporting running also
> on
> >> a KVM host. So far our Vagrantfiles and boxes are virtualbox-specific.
> >>
> >> Cheers,
> >> Dave
> >>
> >>>
> >>> On 23 April 2014 11:24, Rajesh Battala <rajesh.battala@citrix.com>
> >> wrote:
> >>>> Congrats  Ian, Darren and Seif!
> >>>>
> >>>>
> >>>> -----Original Message-----
> >>>> From: Sebastien Goasguen [mailto:runseb@gmail.com]
> >>>> Sent: Wednesday, April 23, 2014 1:43 AM
> >>>> To: dev@cloudstack.apache.org
> >>>> Cc: Ian Duffy; Darren Brogan; Seif Eddine Jemli
> >>>> Subject: [GSOC2014] Acceptance of three projects for CloudStack
> >>>>
> >>>> I am pleased to announce the acceptance of three google summer of code
> >> projects for Apache CloudStack in summer 2014.
> >>>>
> >>>> Our three students are:
> >>>>
> >>>> -Ian Duffy (back for one more round), who will work on a
> >> vagrant/puppet/travis configuration to improve on devcloud.
> >>>> -Darren Brogam, who will improve the GCE and AWS interfaces (that has
> >> he has co-authored) -Seif Eddine Jemli, who will develop a new UI for
> >> Primary Storage plug-ins
> >>>>
> >>>> Mike Tutkowski will mentor Seif, while I will mentor Ian and Darren.
> >>>>
> >>>> Join me in congratulating our three stars of gsoc 2014 !
> >>>>
> >>>> -Sebastien
> >>
> >>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message