[openstack-dev] A few questions on using COE puppet/cobbler...

Paul Michali pcm at cisco.com
Mon May 20 15:23:08 UTC 2013


On May 20, 2013, at 11:00 AM, Mark T. Voelker wrote:

>> *Q: Any idea as to what I'm doing wrong on the proxy setup here?*
> 
> Is it happening consistently?  We've often seen Cisco's internal proxies
> fail intermittently, so if you've only tried it once I wouldn't be
> surprised if it worked again next time around.

PCM: Yeah, and actually, "pip install collectd" failed too.  I just tried creating a pip.conf file, with a longer timeout and using a different index:

[global]
timeout = 180
index-url = http://g.pypi.python.org/simple
download-cache = ~/.pip/cache

Now it says:

err: /Stage[main]/Collectd/Package[collectd ]/ensure: change from absent to present failed: Execution of '/usr/bin/pip install -q --index-url=http://ucs-build-server/packages/simple/ collectd ' returned 1:   Could not find any downloads that satisfy the requirement collectd
No distributions at all found for collectd
Storing complete log in /root/.pip/pip.log

Maybe that repo is not good to use. I could retry with a different repo.


> 
> Note that collectd is being installed via pip rather than apt, so the
> $location actually doesn't come into play here.  That also means that
> you need to be able to reach the public internet, of course.

PCM: The other pip installs worked, so I have access in general.  I'll try w/just the timeout.  Any other ideas?


> 
>> *Q: Will we run into an issue with the IP addresses for the management
>> interfaces given we'll have two build servers, each with DHCP servers
>> handling the PXE boots?*
> 
> It would be far simpler not to have multiple DHCP servers serving the
> same network (consider segregating them on separate VLANs), but this can
> work.  You mostly just have to make sure that each DHCP server only
> responds to requests for the hosts it's "supposed to".  Still, it's
> probably going to make life harder than it could be.
> 
>> *Q: Should we create subnets and partition up the space?*
> 
> That would probably make your life easier.


PCM: So maybe use 14.1.0.10-.19 for the HP servers when we re-IP their management IP addresses?

The thought, BTW, was to use the lowest part of the host IP to indicate the system, so that power IP, management IP, and host name all relate (UCS using 10-19, Eclipse 20-29, and HP 30-39). We also will allocate a block of ten VLANs, going from # * 10 to # * 10 + 9, to each node for their data/private network, so that they have exclusive VLANs for use on the Nexus switch.

Regards,

PCM

> 
> At Your Service,
> 
> Mark T. Voelker
> Systems Development Unit
> +1 919 392-4326
> 
> On 05/20/2013 10:43 AM, Paul Michali wrote:
>> Hi!
>> 
>> We're trying to setup multiple build servers (in VMs) in the lab, so
>> that we can automatically provision different types of hardware (have HP
>> Pro Liant, Eclipse, and now UCS boxes). With the current setup we have
>> an operational build server that provisions the HP Pro Liants with
>> either COE or Devstack. Works well.  Here are the questions, related to
>> the second build server VM that we are setting up (and later will do a
>> third for the Eclipse boxes):
>> 
>> With this build server, I ran puppet apply and am getting this error:
>> 
>> err: /Stage[main]//Node[master-node]/Exec[pip-cache]/returns: change
>> from notrun to 0 failed: /usr/bin/env
>> http_proxy=http://proxy-wsa.esl.cisco.com:80
>> https_proxy=http://proxy-wsa.esl.cisco.com:80 /usr/local/bin/pip2pi
>> /var/www/packages collectd xenapi django-tagging graphite-web carbon
>> whisper returned 1 instead of one of [0] at
>> /etc/puppet/manifests/core.pp:421
>> 
>> It appears to be a proxy issue, but I'm not sure what is wrong as this
>> build server has the same thing as the other (working build server).
>> The site.pp has:
>> 
>> $proxy                  = "http://proxy-wsa.esl.cisco.com:80"
>> $location               = "http://128.107.252.163/openstack/cisco"
>> 
>> The /etc/apt/sources.list.d/cisco-openstack-mirror_folsom has:
>> 
>> # cisco-openstack-mirror_folsom
>> deb http://128.107.252.163/openstack/cisco folsom main
>> deb-src http://128.107.252.163/openstack/cisco folsom main
>> 
>> This command fails with a timeout, when run manually as well.
>> 
>> *Q: Any idea as to what I'm doing wrong on the proxy setup here?*
>> 
>> 
>> The second question relates to idea that we want both of these build
>> server VMs running at the same time. The build servers are on the
>> 192.168.220.0/24 network. The UCS boxes will have power management IPs
>> on 13.0.0.0/16 and management IP will be on 14.0.0.0/16 network, using
>> host part of .30 to .39 for the ten systems.
>> 
>> Currently, the HP boxes power and management ports are on the
>> 192.168.220.0 network, but the intent is to move these to the 13.0.0.0
>> and 14.0.0.0 networks as well using host IP parts (.10 to .19). In the
>> future, we'll alter the Exclipe boxes too, to use .20 to .29 in these
>> same IP ranges (only we'll need to use a managed APS for these).
>> 
>> We've statically set the IPs for the power management ports and will
>> rely on the MAC addresses to assign the IPs via cobbler/puppet.
>> 
>> *Q: Will we run into an issue with the IP addresses for the management
>> interfaces given we'll have two build servers, each with DHCP servers
>> handling the PXE boots?*
>> *
>> *
>> *Q: Should we create subnets and partition up the space?*
>> *
>> *
>> *Q: Any other issues that you see with this plan?*
>> *
>> *
>> *
>> *
>> Thanks in advance!
>> 
>> 
>> PCM (Paul Michali)
>> 
>> Contact info for Cisco users http://twiki.cisco.com/Main/pcm
>> 
>> 
>> 
>> 
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>> 
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 495 bytes
Desc: Message signed with OpenPGP using GPGMail
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130520/d6869f16/attachment.pgp>


More information about the OpenStack-dev mailing list