Everything is a Freaking DNS problem

So 2006 again will be yet another year where I`ll mis the Xen User BOF at Ottawa Linux Symposium which Ian just announced and
the rest of OLS :(

(15:42:23) Snake007uk: guys i have this problem, if i (ssh via putty) onto a newly built box with or without pubkey after i enter username i have to wait like 30secs for passsword prompt to appear, the same happens once i login if i su -
(15:42:31) Snake007uk: any ideas
(15:42:32) orc_orc: Snake007uk: dns issue
(15:42:34) Evolution: Snake007uk: check dns
(15:42:46) orc_orc: (rimshot)
(15:43:01) Snake007uk: why dns ? just curious ?
(15:43:10) Zathrus: Snake007uk: reverse lookup on you
(15:43:17) orc_orc: Snake007uk: ssh does reverse lookupos -- something is horked in yours
(15:43:25) Evolution: Snake007uk: because ssh does a reverse lookup to see that you're coming from where you say you are.
(15:43:29) Evolution: security and all that.
(15:44:17) Snake007uk: the dns is fine...  i can ping out anywhere i want
(15:44:26) Evolution: Snake007uk: no, it's not.
(15:44:28) orc_orc: Snake007uk: _reverse_ lookups
(15:44:38) ***sdog wonders why these t-shirts aren't selling good ..http://www.spreadshirt.net/shop.php?sid=79281
(15:45:02) orc_orc: sdog: indeed
(15:45:08) orc_orc: but really only 80%
(15:45:11) orc_orc: the rest is women

pdx6 already mentionned last week but the Xen Enterprise Beta program has started, actually this is already the second beta round , the first one was only open to a really limited number of people.
so second experiences coz back when the first beta came out in early april I already had my first experiences... :)

Anyway ... with Novell and RedHat claiming Xen support in their future products, with opensource projects starting to try managing virtual machines, XenSource is up against some heavy competition, But from what I`ve seen so far, they are ahead of the competition. Xen-Enterprise is rather easy to install , as it isn't a package you need to install on an already installed Xen platform with a zillion non trivial dependencies but a complete distribution itselve, so you just pop in the CD and install it. Apart from the dedicated server (actually it's not so dedicated as I installed it on a sepearte partition of a box already running Xen) it also has a management client that you use from a remote platform which does not need to run Xen, so this can be your laptop or any other workstation. (The CD comes with both RPM's and some weird .exe thingie) from that management client you can manage multiple Xen Enterprise servers.

Apart from multiple servers you can also run multiple Distributions, within the previous beta's there was only support for Debian and RHEL but now Xensource added support for Suse (which I still have to test), installing both Debian and RedHat goes smooth from their gui (smother than the other alternatives I've seen) and stopping, suspending and restarting machines is also a feature where the competition is less succesful.
It's still a Beta and obviously there are things that can use some improvement, but Xensource is listening to it's beta testers, just the gui alone has improved a lot from it's previous beta's. So I`m sure they are heading the right direction.

With much delay , I finally found some time to write down my first experiences and toughts on openQRM, I've ben working with Matt of Qlusters doing lots of tests for different plugins and features of openQRM. And what can I say.. I love the tool, but I won't be putting it into production yet at a customer yet.

openQRM today expects you to boot a full operational system over the network, while it has some advantages such as being able to boot the same identical image on different hardware with a couple of mouse clicks, I still prefer to deploy
an image physically on a disk, certainly in the customer environments I'm facing today that method is preferred , but within an environment where servers stay at the same location and networkwise fairly close to a central server and when they are not being shipped to the other side of the world openQRM surely is an option :)

Within that filosophy of booting over the network openQRM does a great job giving people a nice webconsole from where they can mange and deploy new nodes and with their nice plugin system tools such as webmin and Nagios are easily integrated in one management frontend. (Did I mention you don't need to use the web frontend, you can use command line interfaces also.. )

Now the really good news is that I`m not alone with my request for deploying images rather than booting images, and that I've already been discussing with the openQRM team at LinuxTag on how I would do such a thing.
openQRM gives a new meaning to the word Virtualisation, it gives you isolation between the hardware and the software platform and because you don't need to be aware where you have deployed a certain resource it doesn't really matter anymore if it is virtual of physical.

I`m about to take a closer look at how partitioning works in openQRM... I`ll be back with more toughts about it :)

Geert and I decided it was time for a MySQL Usergroup in Belgium , actually mainly a meeting so we set a date and I`m opening our office for the first meeting.

More info: First MySQL Usergroup Belgium at X-Tend (Tuesday, August 29, 2006)

This OLPC Board

was on my desk for the full 30 seconds till Philip took it back to play with it :)

I ran into XenMan a couple of days ago it seems to require Xen 3.0.2
Needs FC5 and Xen 3.0.2 so I upgraded my running Xen platform on FC5.

Altough it looks nice at first, you can easily install new FC5 instances from within the application, A couple of annoyanies occured, it seems like Xenman
can't parse multiline disk configs that are no problem for Xen itselve.
I also can't find LVM support which is crucial., so I`m stuck with large loopback files. Their Create Domain option screen also expects me to work with multi
G disks , where as in a typical Virtual Machine environment I tend to keep my disk as small as possible, partitions of a couple of hundred megs really aren't exceptional

Sometimes an error compeletely aborts the program

Error: I need 129 MiB, but dom0_min_mem is 256 and shrinking to 256 MiB would leave only 37 MiB free.

Is a fair error, but even when this error occurs you should be able to other actions from the gui, not having to restart it again.

Oh when one of those installs fails Xenman doesn't clean up after itselve. (:() Upon retrying with more memory available, it works, it quickly created
a virtual machine that starts installing Fedora.
In my first attempt it stopped , probably because I only gave it 64 Mb to play
with.
An installation of FC5 eventually succeeded, however booting the actuall image was yet another issue. At first it started booting the installer again.

I filed a couple of bugs to their sf.net account and they seem fairly repsonsive to them .. so this might become a really good product in the future... one to keep an eye on..

Just a small note that X-Tend is accepting internship applications, if you are a student with an interrest in opensource you might want to do your stage with us.
We have different projects available ranging from system management automatisation to gui development.

Just contact me if you are interested .

PS. Totally unrelated , I`m also looking for a junior java developer

Whurley just pointed to Adventures in Open Source : Open Source is not a Marketing Term on the OMC Mailing list.

This article has some really really good points that I have also been explaining to different folks before, and that I want to focus on again .. they are mainly about what open source is not .. and which kind of wanna be opensource companies or porjects therefore you should better not waste your time on .

Let's have a look :)

Open source software development is not just about providing the source code for your application. It is much more about building a community around a shared project. That takes time. I think the biggest myth about open source software is that you say “hey, I’m open source now” and suddenly thousands of qualified people give up nights and weekends to work on your code.

Or in other words, Open Sourcing your end of life products helps noone.

He then compares Open Source to carpenting

You can buy the same tools that a master craftsman uses, but it is doubtful that you can produce the same quality of work without investing a lot of time.

That's probably only a part of the comparison as carpenters don't make their own tools or enhance the tools of their collegues. But if they made their own tools, which one would you choose, the one that made , enhanced it or wrote the book about his tools, or the one that just started experimenting with a tool he never used because you as a customer tought it might be interrested ?

There is an independent community supporting the project.

This is really crucial , every project starts with one or more people writing code, but if that group doesn't grow and the project doesn't evolve then what's the difference between binaries you get and a real open source project with onctributions, indeed the lack of peer review.

The company touting an open source application actually had a hand in writing it.
Most of my loathing is reserved for those companies that claim to be open source, but in fact all they did is add a proprietary front end to an open source application. Ethan Galsteed’s Nagios has been a victim of this where his work and the work of his community has been co-opted by “parasites” who hope to cash in on the “open source” buzzword. They also tend to fail at the first two checks in the list as well.

Oh yes Nagios has been a victim of these a lot, lots of companies out there that have written multiple plugins and are still selling those as proprietary software also ..luckily not all of them claim to be open source companies.
Let's just hope they learn and contribute back some day ..

As a reply to the proposal MichaelDehaan from RedHat made on an OpenManagement Backend Format made

Jamie Cameron of Webmin fame writes

That looks like an excellent proposal to me .. I plan to start adding an XML interface to Webmin in the near future that will allow its existing backend API to be accessed through calls like those you described.

Now that's a really good direction ...

I used to dislike webmin because it gave incompetent users a reason not to search for help because they could just click away and destroy their systems from their browsers (freel free to s/webmin/yast/g;) , so I never really looked closer a it , I've also always been under the impression that webmin just didn't scale, if you wanted to modify something on 5 machines you had point your browser to 5 different urls.

If however you have an interface to webmin , and those are being standardised, you could actually centrally tell a group of machines to do a certain action. Still no CFengine or Puppet replacement but probably quite usefull for a large group of people.