problem with openstack-install

[Jeff McLamb]

Hi Adam -

Thanks for the info and the quick response! I really appreciate it.

Just to be clear, the proxy that I speak of is the maas-proxy that was
installed when I installed maas as per the instructions here:

http://www.ubuntu.com/download/cloud/install-ubuntu-openstack

I do not actually have a proxy server of my own, but I figured the
proper setup would be for all of the servers that enlist in MAAS to
get their DHCP and DNS from a MAAS-managed interface. There is nothing
requiring this proxy server. The MAAS server is the default gateway of
all of the PXE-booted servers with ip_forwarding enabled in sysctl and
the proper MASQUERADING rules applied in iptables, so I wonder if I
could just get rid of the maas-proxy altogether?

Alternatively, I could set it up such that each of the enlisted
servers in MAAS has one of its non-PXE interfaces in a VLAN that gets
a direct Internet connection via DHCP from the ISP, although I don’t
know if the MAAS nodes with automatically enable and get a lease on an
interfaces that is not its PXE-booted interface, once they are
commissioned.

By all means feel free to share anything on the troubleshooting page.

Thanks,

Jeff

On Tue, Jul 28, 2015 at 6:43 PM, Adam Stokes <adam.stokes at canonical.com> wrote:
> Hi Jeff,
>
> So proxy support within Landscape Autopilot is being looked into for
> deploying OpenStack within the autopilot. Currently the only way to have
> proxy support is to use the Multi install.
>
> As for multiple juju bootstraps that is expected as the environment housing
> landscape and its service dependencies is a separate environment from which
> openstack and it's service dependencies are deployed.
>
> I will ping the landscape guys to have a look at this email and give some
> more input on the possibility of proxy support within Autopilot.
>
> Also wrt some of your findings do you mind if I use those points on our
> troubleshooting page? (I'm currently working on getting a new site up with
> additional resources)
>
> Thanks
> Adam
>
> On Tue, Jul 28, 2015 at 6:16 PM, Jeff McLamb <mclamb at gmail.com> wrote:
>>
>> Hi everyone, I just joined the list and in perusing the archives I
>> found a thread from June about issues bootstrapping Juju because of
>> failed tools downloading.
>>
>> I finally managed to get beyond this issue and wanted to offer my
>> solution:
>>
>> In addition to adding the appropriate http-proxy and https-proxy lines
>> to the environments.yaml file, found that I had to make sure the
>> MAAS/Juju server had the same timezone configuration as the bootstrap
>> node. Turns out one was set to EDT and one was UTC. When I issued a
>> 'sudo dpkg-reconfigure tzdata' on the MAAS/Juju server where I run
>> openstack-installer, it successfully downloaded the tools and
>> completed the installation, giving me a Landscape web UI running in an
>> LXC container within the bootstrap VM.
>>
>> To give a bit of background about my setup, I have 4 servers in MAAS
>> that are typical big-RAM big-core machines, power-controlled via PXE.
>> So that I would not have to waste one of them for a
>> bootstrap/Landscape node, I created a KVM VM on the MAAS server that
>> PXE booted into MAAS, and commissioned it. When I ran:
>>
>> openstack-installer --upstream-ppa --http-proxy <maas-proxy:8000>
>> --https-proxy <maas-proxy:8000>
>>
>> It chose the VM as the bootstrap node as I was hoping (although I did
>> not direct it to do this, just my luck?), and installed Landscape and
>> all of its charm dependencies in 5 different LXC containers on the VM.
>>
>> I am running MAAS on a fully update 15.04 server, using the
>> experimental ppa for cloud-installer and the latest juju from the
>> proposed ppa.
>>
>> Now that I can login to Landscape and I have 4 available servers
>> listed on the OpenStack (Beta) tab, I can choose my deployment options
>> (KVM, Ceph, etc.) and then I select all 4 servers, even though only 3
>> are required. This is a new reduced minimum from the previous 5 in
>> this latest version of LDS.
>>
>> When I click on install, the first step it tries to do is to bootstrap
>> Juju! It chooses one of my physical servers in MAAS and tries to
>> bootstrap. However, it never even gets to powering on the server, as
>> it eventually stalls out with the following error:
>>
>> juju ended with exit code 1 (out='', err='Bootstrapping environment "4"
>> Bootstrap failed, destroying environment
>> ERROR failed to bootstrap environment: Juju cannot bootstrap because
>> no tools are available for your environment.
>> You may want to use the 'agent-metadata-url' configuration setting to
>> specify the tools location.’)
>>
>> Why would Landscape be trying to re-bootstrap a new Juju environment?
>> Shouldn’t it just use the one I already used to deploy Landscape
>> itself?
>>
>> I logged into the landscape/0 unit from the MAAS server with the
>> following command:
>>
>> $ juju ssh landscape/0 sudo
>> ‘JUJU_HOME=/var/lib/landscape/juju-homes/`sudo ls -rt
>> /var/lib/landscape/juju-homes/ | tail -1` sudo -u landscape -E bash’
>>
>> From here it takes me to the “4” environment (representing my latest
>> attempt, there are other directories for my previous attempts with
>> 1,2,3)
>>
>> The environment does have an environment.yaml file that points to my
>> MAAS server, but it does not have any http-proxy or https-proxy stuff,
>> and is not near as complete as the one on my MAAS server.
>>
>> This is where I am currently stuck. There is also the Juju tab in the
>> Landscape UI where I manually added the Juju environment details for
>> the one on the MAAS server. It is listed as a valid Juju environment
>> with a valid endpoint, but it seems not to care to use this one when I
>> try to deploy OpenStack.
>>
>> Any help would be greatly appreciated!
>>
>> Thanks,
>>
>> Jeff
>>
>> --
>> ubuntu-openstack-installer mailing list
>> ubuntu-openstack-installer at lists.ubuntu.com
>> Modify settings or unsubscribe at:
>> https://lists.ubuntu.com/mailman/listinfo/ubuntu-openstack-installer
>
>