Some readers trying out the module have noted missing packages. I completely missed this since I was developing with a system with git packages. We’ve released razor module version 0.1.1 addressing this issue. So if you have a problem try again, and feel free to contact me for any other issues.

Seeing node and mongo make their way in to infrastructure is distressing.

Lars

It’s open source and the persist layer is pluggable. It is easy to replace mongo with something else if you prefer.

https://github.com/puppetlabs/Razor/tree/master/lib/project_razor/persist

Hey Lars,

Curious about the Node.js & Mongo comment. Not that I disagree, more that I want to understand.

What would you prefer over Mongo? And what would you prefer over Node.js – that would give comparable event-driven performance?

.nick

What exactly are we supposed to do for the following step:

On the puppet master assign the razor class to the appropriate node, and trigger a puppet agent run afterwards

I already have a DHCP server running and I’m currently just pointing next-server to Razor system. I’ve already loaded the micro boot image in Razor but not sure what this particular step is referring to.

Thanks

–William

William,

There’s two ways to execute puppet. In client server mode, or apply the puppet manifests locally.

If you only have a single server, install puppet, download the modules and run ‘puppet apply’ to deploy the application. When it’s done move on to using Razor. Most Linux distros include puppet packages, but they are not always up to date. The enterprise version is available at: http://puppetlabs.com/download/, or use apt.puppetlabs.com for the latest open source version on Ubuntu.

If you have puppet deployed in your environment. On the puppet master, download the modules, then in /etc/puppet/manifests/site.pp specify which node would deploy the razor application (in the blog the example node name is ‘razor_host’) . When the puppet agent runs on the client system it will deploy razor and report the results back to the puppet master.

In master/agent mode depending on what other components are deployed, you can specify the host configuration via other methods, and trigger the puppet agent using mcollective via CLI or in Puppet Console GUI.

Thanks,

Nan

Nan,

Currently I have a single system which is running puppet + razor (following the guide). I have an existing DHCP server which I’m forwarding the request over to puppet/razor system which from what I can tell has an active TFTP server running which is hosting the MK image. I’m able to boot a blank server which gets the right IP Address and forwards the request using next-server stanza over to puppet/razor system. It performs the iPXE but throws the following error:

DHCP (net0 00:50:56…)….ok
http://razor:8026/razor/api/boot?mac=00%3A50……connection reset (ipxe.org/err/0f0a6039)
Could not boot ttp://razor:8026/razor/api/boot?mac=00%3A50…
Operating System not found

I’m assuming the boot process works just like with any traditional PXE install w/DHCP+TFTP, but I’m not familiar with puppet, so not sure if there was something else I needed to do.

Let me know if there’s a particular step I still need to perform or another system I may need to deploy?

Thanks

Are the node web service running? By default they should be listening on port 8026, 8027.

$ /opt/razor/bin/razor_daemon.rb status
razor_daemon: running [pid 3721]

$ ps -ef | grep node
root      3732  3721  8 17:27 ?        00:00:00 /usr/bin/node /opt/razor/bin/api.js
root      3736  3721 10 17:27 ?        00:00:00 /usr/bin/node /opt/razor/bin/image_svc.js

Check the log file in log/razor_daemon.log and see if there are any errors.

By default razor should boot the microkernel, and the initial boot request should respond with a valid ipxe boot file (using arbitrary example mac address):

$ razor -w boot default '{"mac":"00:0c:29:f7:4c:1d"}'

#!ipxe
kernel http://192.168.232.141:8027/razor/image/mk/kernel || goto error
initrd http://192.168.232.141:8027/razor/image/mk/initrd || goto error
boot || goto error

:error
echo ERROR, will reboot in 5
sleep 5
reboot

The kernel and initrd address should be valid. You should be able to wget both files, if not check the output of razor config, and update conf/razor_server.conf with the valid ipaddress/port.

Just wanted to note that after doing a clean install with Ubuntu Precise server in a Fusion 4 VM I had to manually kick off the web service. It does not get started after applying the razor module to the local node. Once that was done PXE boot completes successfully and new nodes are registered. On to adding my first Ubuntu Precise provisioning template.

Pepijn,

Thanks for the info. I have updated the blog post to include the service status check and modified the puppet deployment module. I have a pull request that will start the service automatically. Once it’s reviewed and merged, I will publish an update that resolves this.

Thanks,

Nan

Nan,

Okay, looks like that was my problem as mentioned by Pepijin, the web service was not running. Once I got it running, then the request which was forwarded from my DHCP server immediately booted up the MK image as I expected.

Thanks

I noticed while creating a model for ESXi 5, it does not allow you specify Datacenter and Cluster names with a “-” (e.g. cluster-01).

Also a general question, is there a way to get a list of all the options for each slice? I’ve found that using –help|help|-h does not really provide enough details and it’s only from both Nick’s and this post that I’ve been able to identify some of the commands.

Thanks

We did some rewiring of slices, and the CLI help is step behind some of the changes. On the github page there’s a wiki where we have documented most of the slices: https://github.com/puppetlabs/Razor/wiki

The wiki should be open for editing, so by all means feel free to contribute, and if you feel like getting your hands dirty, dive under the hood and look at the code in lib/project_razor/slices/.

The cluster name and datacenter name is a regex validation, just file an issue on github and we can change it to support -. We will handoff the ESX node a puppet master, and in puppet you would express where this is deployed, so we might make some changes to the model. I have follow up blog post discussing how broker handoff is done both for puppet agents as well as ESX nodes, so stay posted.

[...] This is very much cribbed from the Puppet Labs page here [...]

I’ve setup everything per this guide after trying to do it manually myself and get the same issue. Everything works and the test VM’s I have up boot up and are shown in razor node, and the policy matches, but they never kick off installing…What could I be missing?

Discovered Nodes
UUID Last Checkin Status Tags
4Ql3igAWAnpx9xCkW0lmfm 9 sec A [IntelCorporation,vmware_vm,nics_1,cpus_2,memsize_1GiB]
6lc736pDwpTAKCvDhK3RGI 49 sec A [IntelCorporation,vmware_vm,nics_1,cpus_2,memsize_2GiB]
root@Razor:~# razor policy
Policies
# Enabled Label Tags Model Label Count UUID
0 true precise [tags=IntelCorporation,vmware_vm,nics_1,cpus_2,memsize_1GiB] install_precise 0 5IUfNPcU2XJqvi4gfaBoec

Jason, please file an issue on github: https://github.com/puppetlabs/Razor/issues

Include the information above and the output of ‘razor policy active’ and ‘razor policy active log’.

I am trying to get Razor up and running but have hit a roadblock. I have a fresh Ubuntu install of precise with puppet/puppetmaster. I did the razor install through puppet, started the DB and the node hosts (8026/8027). After downloading the mk image and adding it to Razor I configured a test client with the appropriate options on my DHCP server (MS).

The client will PXE boot and it contacts the server running razor and loads the menu. Then it transfers ipxe.lkrn fine and errors out with the following:
“Could not fetch razor.ipxe: Operation not supported (http://ipxe.org/3c092003)
Operating System not found”

I did a packet capture from the razor host and I never see a request for the transfer of the razor.ipxe file after ipxe.lkrn. I’m not sure what to try next. I didn’t see anything in the comments above or on the wiki/issues list.

Both client and server are running on ESXi v5. I tried configuring the client with the VMXNET3 adapter and the E1000, both had the same result.

It’s possible razor config auto detected the wrong IP address of the server. What’s the output of?

razor -w boot default '{"mac":"00:00:00:00:00:00"}'

You should get output similar to (your IP will be different):

#!ipxe
kernel http://192.168.232.141:8027/razor/image/mk/kernel || goto error
initrd http://192.168.232.141:8027/razor/image/mk/initrd || goto error
boot || goto error

If you check, can you fetch the kernel and initrd file remotely?

wget http://192.168.232.141:8027/razor/image/mk/kernel

Also I would encourage usage of github issues and wiki: https://github.com/puppetlabs/Razor/issues. Please send relevant output from log/*.log as well if the info above isn’t sufficient to resolve your issue.

Thanks,

Nan

Hi Nan,

We are trying to get razor up and running. When running this command “puppet apply /etc/puppet/modules/razor/tests/init.pp – -verbose”, we got the following error:

err: /Stage[main]/Mongodb/Apt::Source[10gen]/Apt::Key[Add key: 7F0CEB10 from Apt::Source 10gen]/Exec[b94ff96203335a23dcd698a0ca82d59bdc9b5292]/returns: change from notrun to 0 failed: apt-key adv –keyserver ‘keyserver.ubuntu.com’ –recv-keys ’7F0CEB10′ returned 2 instead of one of [0] at /etc/puppet/modules/apt/manifests/key.pp:53

Do you know how to get around this?

Thanks

Xing:

I’m assuming you are running puppet apply with root permission. Can you run the command as root and provide the error message?

apt-key adv --keyserver 'keyserver.ubuntu.com' --recv-keys '7F0CEB10' 

Thanks,

Nan

Nan,

Yes, I’m running as root. Here’s the message:

~# apt-key adv –keyserver ‘keyserver.ubuntu.com’ –recv-keys ’7F0CEB10′
Executing: gpg –ignore-time-conflict –no-options –no-default-keyring –secret-keyring /tmp/tmp.UuZJvgn6zt –trustdb-name /etc/apt/trustdb.gpg –keyring /etc/apt/trusted.gpg –primary-keyring /etc/apt/trusted.gpg –keyring /etc/apt/trusted.gpg.d//pl-enterprise-extras-keyring.gpg –keyring /etc/apt/trusted.gpg.d//pl-keyring.gpg –keyserver keyserver.ubuntu.com –recv-keys 7F0CEB10
gpg: requesting key 7F0CEB10 from hkp server keyserver.ubuntu.com

It’s hanging while requesting key. It could be that our IT is blocking this. Is there a way to get around this?

Thanks,
Xing

Nan,

We retrieved the key using a different port which is not blocked. Then we installed mongodb manually and was able to finish the install.

Thanks,
Xing

How can we reprovision a machine? I had a number of machines setup under one policy, which didn’t have a puppet broker. Now I want to switch those machines to a different policy. I created the new policy, but when I boot the machines they end booting locally instead of having the new policy applied to them (new OS install + puppet broker).

Thanks.
J

Jack, first identify the node’s active model binding via:

razor policy active
razor policy active log {uuid}

Then remove that system active model (which was bound by policy rules):

razor policy remove active {uuid}

p.s for anyone having an issue related to MarkC comment see:
https://github.com/puppetlabs/puppetlabs-razor/issues/13

Thanks Xing. Is there a way to ‘unbind’ a node from a policy? I don’t want to delete the whole policy, and I don’t want a different policy applied; I’m in trial/error mode, and so would like to re-provision the machine trying different settings?

Thanks.
J

I think you want to disable a policy, otherwise if a node hardware facts matches it will get rebound to the same policy.

razor policy disable {uuid}

Again unbind a node’s active model is:

razor policy remove active {uuid}

This is different from delete policy which is:

razor policy remove {uuid}

Thanks,

Nan

Would that allow me to re-provision the machine? That’s what I’m trying to achieve; go through the re-install/commissioning process again…

Thanks.
J.

Yes, that will allow you to install another OS based on the current policy.

Hi Nan,

I am in a similar situation as Jack. I need to ‘unbind’ an active policy. Currently I have removed all the policies, models and images, in order to try and re-provision a node again from the beginning. However, there still seems to be an ‘active’ policy when I do a ‘razor policy active’. How can I remove this active policy?

Thanks,
Swapnil

Swapnil, since there still seems to be some confusion, I’ll elaborate on the answer above and give the output as well:

1. identify the node you want to unbind.

$ razor policy active 
Active Models:
    Label          State           Node UUID         System  Bind #           UUID           
install_centos  os_complete  74hlvTYRrBtwtEuZXnHXkg  puppet  2       7HmohnvcR4lxFOakxoukqg  
install_centos  os_complete  XNnfl9IzvB0FUfnjryGbq   puppet  3       fktzuQfL8o8oVeoj7ciLK   
install_centos  postinstall  3qBqWhuIXh8gEMHz7aBK7m  puppet  1       41ecNNJzSvlI5MhdRhoa1i 

2. verify this is indeed the node you want to unbind.

$ razor node 3qBqWhuIXh8gEMHz7aBK7m
 UUID =>  3qBqWhuIXh8gEMHz7aBK7m
 Last Checkin =>  05-31-12 13:32:25
 Status =>  bound
 Tags =>  [cpus_1,IntelCorporation,memsize_1GiB,nics_1,vmware_vm]
 Hardware IDs =>  [000C29AD79F0]

3. remove that node active model (which was bound by policy rules):

$ razor policy remove active 41ecNNJzSvlI5MhdRhoa1i 
Policy remove_active
 Active Model [41ecNNJzSvlI5MhdRhoa1i] removed

4. this node will no longer show up under razor policy active

$ razor policy active 
Active Models:
    Label          State           Node UUID         System  Bind #           UUID           
install_centos  broker_fail  74hlvTYRrBtwtEuZXnHXkg  puppet  2       7HmohnvcR4lxFOakxoukqg  
install_centos  broker_fail  XNnfl9IzvB0FUfnjryGbq   puppet  3       fktzuQfL8o8oVeoj7ciLK   

At this point, reboot the node, it will reregister and razor will provision the system based on your current active policy.

I’m getting Connection Timed out when PXE booting a server to razor. I’m running razor version 0.1.6 and mk version 0.8.9.0, I have dhcpd running on a seperate server. The error is http://ipxe/4c0a6035 and this is what I see from the console:

Loading ipxe.lkrn…
Loading razor.ipxe…ready.
iPXE initialising devices…Ethernet addr: 00:19:b9:d3:0d:4a
Broadcom NetXtreme II (B2) PCI-X 64-bit 133MHz
NIC Link is Up, 100 Mbps full duplex
WARNING: Using legacy NIC wrapper on 00:19:b9:d3:0d:4a
ok

iPXE 1.0.0+ — Open Source Netowrk Boot Firmware — http://ipxe.org
Feature: VLAN HTTP iSCSI DNS TFTP AoE SRP bzImage COMBOOT ELF MBOOT PXE PXEXT DHCP (net0 00:19:b9:d3:d0:4a)…… ok
http:///razor/api/boot?mac=00%3A19&3Ab9%3Ad3%3Ad0%3A4a………… Connection timed out (http://ipxe.org/4c0a6035)
Could not boot http:///razor/api/boot?mac=00:19:b9:d3:d0:4a: Connection timed out (http://ipxe.org/4c0a6035)

I can reach both port 8026 and 8027 from from machince within the same vlan as the server that is PXE booting and when I go to the url in the error I get:

#!ipxe kernel http:///razor/image/mk/kernel || goto error initrd http:///razor/image/mk/initrd || goto error boot || goto error :error echo ERROR, will reboot in 60 sleep 60 reboot

and if I go to those urls I can download the vmlinuz and gore.gz respectively.

Any ideas on whats causing this?

thanks,
-tj

Just noticed that the url in my error got clipped but you get the idea. its going to the ip of my razor server over port 8026.

That doesn’t seem right, you should have an IP address and port number:

kernel http://{ip_address}:{port}/razor/api/boot?

Your output only shows:

kernel http:///razor/image/mk/kernel

Check the output of razor config.

Also please raise issues at the links below:
https://github.com/puppetlabs/Razor/issues
https://github.com/puppetlabs/puppetlabs-razor/issues

I’m going to document the known issues under:
https://github.com/puppetlabs/Razor/wiki

Since the comments can get out of date because of updates to razor and the razor module, I’m going to close the comments for issues.

Nan, I have raised an issue at the link you provided but I wanted to be clear that I have http://(ipaddress:{port}/razor/api/boot? but when I typed it in a used less than and greater than signs and they got stripped out. Sorry for the confusion.

In the DHCP module example at the top of the posts, there’s a missing single-quote at the end of the “‘puppetlabs.lan,” definition. Also, it seems the links to the puppetlabs/dhcp page on Github is not working properly.

Matthias, thanks for corrections, I’ve updated the post.

Hi,

Is there a detailed documentation about the broker handoff and how it works? The above blog says that the broker handoff will be discussed in a follow up article. Can I have the link to that article or any documentation that discusses how the broker handoff works and its purpose etc?

Thanks,
Swapnil

I’m not really satisfied with the current agent handoff implementation, and held off releasing the follow up blog post. I’d much rather discuss this when the feature have been updated, rather than document the current setup.

Thanks,

Nan

[...] http://puppetlabs.com/blog/introducing-razor-a-next-generation-provisioning-solution/ http://puppetlabs.com/blog/puppet-razor-module/ http://nickapedia.com/2012/05/21/lex-parsimoniae-cloud-provisioning-with-a-razor/ [...]

Thanks Nan. I had another doubt here. I actually want to setup and configure several other packages on the new physical host and VM along with the baremetal provisioning. Will this be done by the broker handoff to puppet? If not, how can I install and configure other packages (Eg: puppet) automatically on the new VM after the OS is configured by Razor on it?

Swapnil

Swapnil, the reason I’m not happy with the current implementation is because it installs puppet via gems. I’d much rather this ben done via packages (via yum.puppetlabs.com and apt.puppetlabs.com), but it’s not available universally and require platform specific repos.

The models are intended to be adjustable and you can create additional models with different software packages, or use puppet to install additional software application afterwards. If you are happy with puppet installed via gems, use the current handoff process, otherwise modify the models in lib/project_razor/model and update the appropriate preseed/kickstart erb template to include the appropriate puppet packages and you can use puppet to manage and deploy additional software application.

Again, I’m still looking for a better solution, and I’m looking at replacing puppet gem via puppet node face installation process so it can support puppet using platform specific packages for opensource and Puppet Enterprise.

Nan

[...] http://puppetlabs.com/blog/puppet-razor-module/ [...]

[...] http://puppetlabs.com/blog/puppet-razor-module/ [...]

Nan,
I had a question about the tagging. If we have 10 nodes but only want to identify 2 nodes to be provisioned by Razor, how to identify them? Is this purely done by tagging? Eg: In the above case, will we need to tag the 2 nodes separately under 1 tagname?

Swapnil

[...] part of running the puppetlabs razor module. I am pretty much following along with what they say on http://puppetlabs.com/blog/puppet-razor-module/ where they say the configuration is “Easy”. I say, let me decide for myself if [...]

One thing I noticed is that razor does not have a template for windows bare-metal. Is that something that is possible? I am looking for razor to do this along with RH, ESX, and CentOS.

Hey. So, you are right: we don’t have a template for windows bare-metal installation (or, in fact, Windows installation at all.)

Nick Weaver is the author of that, and has not gotten his code to the stage it can be release yet. Without that, while we would like to add the feature it isn’t at the top of our list.