Ubuntu Cloud Portal: Juju at PuppetConf

Jos is working for Krux Digital, he’s taking care of all the ops with all the servers living in EC2, he talked extensively about Ubuntu Server and how it is the best OS for cloud right now. He made some very  good points about using Vagrant for deploying developer VMs and trying  to design your platform in order to always preview failure since it’ll  happen. They also have a very concise operative structure, designed to avoid common failures at EC2 like network partitioning, and also designed to heal up in most common scenarios to avoid the sysadmin getting woke up at 4am.

You can see his talk at http://www.slideshare.net/jiboumans/oneman-ops-9383061

Gareth  talked extensively about Vagrant and how extremely easy is to create images and be able to deploy them with Vagrant. He touched some advanced  subjects as how to replicate images and keep private images that you  can update in your own timing.

Pedro  works at Facebook Engineering, he explained the best practices they’re  using (they currently use CfEngine) inside Facebook to automate all  their machine deployment and some insight about how Facebook scales  their stuff. They keep all their machines in clusters and do deployments of 5%-10%-95%-100% to be able to roll back  quickly in case there’s an issue or problem with the new code. He also  talked about the common logging stream in a single unified web page of  all the operations made on the systems, both by automatic scripts and by  sysadmins, very useful.

Another  talk about best practices and how do they deploy puppet in their  infrastructure, people are trending towards not having multiple  configuration files with just small changes but rather have one file  that covers all yours needs in order to be able to escalate your  deployment better and not have unreadable puppet configurations.

This  was a very interesting talk, John is the main developer of Noah which is a fork of Zookeeper getting the things he liked most about Zookeeper  into his own program.

Noah takes all the good  parts of Zookeeper (queueing, metadata handling, etc) and gives it an  http interface which is extremely interesting! Have a look at https://github.com/lusis/Noah#readme

You can also follow John in Twitter (@lusis)

Puppet has now created an EC2 instance application deployment tool integrated into puppet so they can deploy instances on the fly and immediately apply a puppet configuration to it.

Also the dashboard got a lot of improvement, the reporting on puppet got a lot better and is now tightly integrated in the dashboard which helps immensely to diagnose any kind of problems you might have.

This  was our talk, we evangelised the crowd about how Juju can be your default application deployment software, and you can expand it using puppet or any other tool or language of your choice!

Update: You can find the slides on http://www.slideshare.net/derleiermann/juju-puppetconf

Martin  is the CEO of Eucalyptus, he explained how Eucalyptus is right now the  most deployed cloud solution in corporations and where they fit in the puppet related world.

This  talk was slightly frustrating, it explained how Google manages their desktop infrastructure (OSX and Ubuntu desktops) using puppet, they did a 10,000 feet overview because being google many details are confidential, nonetheless it confirmed the trend to go towards unified  templates.

Apache  Whirr is a tool to deploy cloud structures, it has some very good input  as the ability to specify memory sizes or other requirements from your  application, Whirr it’ll make sure that it chooses an image size that  matches your requirements properly. All Whirr formulas are written  in Java, you can specify and define a full structure in a java  structured manner. Configs look extremely easy to handle, use standard  config templating that is easy in java, python, ruby, etc. This tool looks very similar to juju in some aspects, definitely worth having an eye on.

This talk covers Whirr but from the ops side rather from the developer side and how they’re deploying whirr at Cloudera.

Chad explains how easy to manage your platform with Whirr and then be able to maintain it with masterless puppet clients.

They use Vagrant to test locally as well.

Very interesting talk about AWS. Highly recommended as soon as it shows up on youtube.

It  started by giving a round-up of of all what Amazon offers to you right  now, separated by layers, very clear and straight forward. He also talked about ways to talk between availability zones. Cloudformation uses a very ruby definition style for creating manifests and cloudformation scripts. Resources  are any resource in Amazon (EC2 instance, balancer, S3 bucket, etc),  you can specify userdata metafiles in your manifests as well. They use clod-init in cloudformation in order to be able to feed data into the instance.

CloudFormer is a new tool from Amazon that can create cloudformation scripts based on your current running resources in AWS.

Very  informal talk! We talked about facter and facter performance, stored configuration in puppet, it was a more QA than anything else.

The  videos will be online soon, the most interesting ones from my point of view, Jos Boumans talk, puppet faces and the AWS talk.

If you have any questions ask away, or leave us a comment right here