Sun, 22 May 2011
VMware vSphere 4.1 HA and DRS deepdive - Short Review
It's been years since I've read a book on VMWare. Between the maturity and ease of use of their GUI tools and my own continual move towards Free virtualisation I've not had the professional need or the spare time to invest but when a book comes as highly recommended as the VMware vSphere 4.1 HA and DRS Technical deepdive does you have to make some room on your (virtual) bookshelf.
Despite its small page count this book covers its subject material in a simple, direct and technically clear way. There is very little fluff and while you could find some of the details buried in VMWare KB articles or white papers its presence here in such a well combined and cohesive form more than justifies the books frankly tiny price tag (at least in the kindle store).
I came away from this book with enough of an understanding of the technologies covered to see where they'd fit, the issues we'd need to monitor for and some of the edge cases that would bite us in deployment. And that's a good return for the small investment of time reading this book takes.
The only downside of the book is that it could really do with another editorial pass or two. While this doesn't alter the quality of the technical content it does make the reading experience a little jarring.
If you want to get in to vSphere HA / DRS then this is a recommended read. Score - 7/10
Sat, 14 May 2011
Wrapping MCollective with Nagios
I've been doing a little tinkering with pre/post release checklists and compliance reporting using cucumber and some Nagios wrapping (among other things) in my test lab and recently needed to do some higher level entire environment checks before moving on to the next step. While it's possible to wrap something like nmaps ping check and then Nagios each target it does feel like stepping back a few years in the tool chain.
Luckily I'm running MCollective, so all this synchronous discovery and polling is in my past. After a little bit of delving in to the existing package and service clients I've come up with a prototype environment wide MCollective backed service check and an MCollective backed package check.
I'm not sure if I'd be willing to replace existing low level checks (for things like cron and ssh processes) with this just yet but it does show how easy it is to wrap MCollective with third party code in order reap its benefits from further down the tool chain. With a little scaffolding hopefully it'll be useful in validating individual policies in security policies and guidelines. But more about that later.
Phase two is probably to pull the scripts together (and just use another parameter to select the resource to check) and to be green or red based on percentage. As an example, requiring 40% of the web servers to be returning 200 before starting the next batch of host upgrades.
ep.io and VMWare at London Devops - May 2011
I never thought I'd use a cliche like "David vs Goliath" but considering the two speakers at London Devops it does seem a little apt. Andrew Godwin from ep.io, a Python hosting platform, was the first speaker, and he did an excellent job of explaining their internal platform, how they make their decisions and what makes them special. While it was both an interesting and engaging talk it did leave me a little worried about the size of the operation.
While small companies are great to deal with in the right situations they can also be a risk due to their low survival odds, questionable ability to grow alongside you and inability to throw resources at an awkward but urgent problem. On the other hand they can provide better levels of support, knowledge and assistance if you can find a good one and treat them more as partners than vendors, and I suspect that ep.io is going to be one of the good ones.
Then we had the VMWare talk. Until a couple of years ago, when budgets shrank again and Xen and KVM began to rise, I was a big fan and a happy user of VMWare products both on server and desktop. While I've not kept up with all the product details it's hard not to have heard of CloudFoundry.
The two speakers, one from RabbitMQ and one from SpringSource (both now part of the VMWare org chart) had very different speaking styles, the speaker from RabbitMQ had a keen wit and kept the tone light with lots of amusing comments like "VMWare is about 9000 staff, about 8000 of them write device drivers" and while the man from SpringSource spent the whole time complaining about how slow his laptop was. At one point the audience nearly had a whip-round to cover the cost of a couple of GB of RAM for him. As for the content it left me a little adrift. I came out of the talk without knowing much more than I went in with. Although I always have to smile when I hear people from SpringSource describe their product line, Spring Tomcat, Spring AMQ, Spring ls and Spring Bash (I might have made the last two up) so it wasn't a complete waste.
Obviously there will be comparisons made between the talk platforms being discussed and one of the most interesting aspects of the evening for me was how well ep.io came out of the deal. They've got an architecture every bit as well thought out as that of VMWares, they're already looking at the next set of problems that both platforms are going to experience and they came across as remarkable professional for such a small team.
CloudFoundry on the other hand will probably have a bigger effect on my working life. VMWare is often quite an easy sell due to its track record and feature set and I can see more companies talking parts of CloudFoundry on board than I can see them hosting with ep.io. So it's one to spend a little time investigating. The fact that it's open source will just make the whole process easier.
The talks were very well attended with 70-80 people in the audience and once again we should say thank you to the Guardian for providing the venue and Gareth for organising it.