funny CLI website theme

Great fun. Via some news link about Lucas' critical post about another new Ubuntu "derivate" (or better: a new product name for an Ubuntu customized for the EEE PC), I got to see a funny/interesting website theme here.
A CLI interfacr to browse a web site.
I still like to view websites by clicking at links etc., but in addition to the CLI, also some of the other themes available for that site are really interesting.
(dunno if it uses a well-known JS lib and that is the one to be cristed, I just like this example...)

Wikipedia Fundraiser, anyone?

Admitted, it took me some time until I decided to give something to the Wikipedia annual Fundraiser. One of the reasons was the usual lazyness, the other was that I waited some days because I remembered that last year, a bunch of big companies had made offers on some days like "we will double all donations made on day X" or in some longer timeframe.

As this didn't seem to happen anymore, I finally spend some Euro's.

Since then, a long time has gone, and I see the number of donations shown on each Wikipedia page to increase only in very minimal numbers each day.

How can that be? Is my assumption wrong that abput the half of internet users are using Wikipedia, a quarter of them on an everyday basis? Are those internet users so poor or too cheap? Did they spend all their money for useless christmas gifts?

I am not really as upset as this sounds now, but I really wonder - I'd think that in about a week, hundred's of thousands of people should have donate at least a little bit... Anyboy reading me, and having an idea?
Are my guessed numbers completely wrong?

As the wikpedia Fundraiser blog suggested, at east I start to do some adveritsing for them now :)
You can do that, too - see here.

NetBSD 4 released - supporting Xen3

The NetBSD has made a new release - version 4.0. This new version is said to support Xen3 dom0 and domU - a.k.a. the system being useable as a host as well as as a guest.

While there where since a long time guest Kernels for NetBSD available which I tested and confirmed to be nicely working on a Linux Xen host, dom0 support was a bit harder to get and use, and I never came to really test it.

Hoepfully, I will find some time to get into it and be able to write a review for this support, and if and how it's compatible to use with Linux and other guest's, soon. You'll read it here.

My first computer

By accident I stumbled upon a picture of my first computer on wikipedia.


An Apple IIe, decommisionned from my fathers engineering office.
It had a green monochromic screen, anwe owned about 2 or 3 games for the system (the onloy one I really remember is "der Fluch das Pharaoh"/"The Pharaoh's Curse", a text/2D graphical role playing adventure game from Synapse Software, released 1983 ) - more could be bought for sums of money absolutely out of the reach of my small pockets from the german apple software distributor pandasoft.

At that time, others had C64's with load of games to play (though most of the pirated) that could be simply connected to a color TV screen. No wonder I lost track with computers after a while, and only got deeply into it about 10 years later :)

Original photo URL - License: GNU Free Documentation License 1.2 or later.

Hamburg at night

When finishing my work yesterday, I took tis nice picture of the old town, seen from Kehrwieder - just wanna share it with you :)

Finally a _real_ use case for Xen and Virtualization

Xen, Vmware, KVM, OpenQRM, SolidICE, XenEnterprise, VirtualIron for things like live migration, HA, software testing, resource management and cosolidation???

No!

We want a virus aquarium!

Technorati claim

I just created a technorati account - to make it believe this blog is mine, here's the claim link:

Technorati Profile

dear lazyweb... two GNOME/Nautilus problems

There are exactly 2 annoying problems, in which I run from time to time, again and again, but I never found a solution, and I didn't make the step to subscribe to some gnome users list to get them solved:

Not always, but quite often when I plug in a usb stick, then navigate through it with Gnome's Nautilus File manager, I cannot unmount it anymore.
When clicking right on it, and selecting unmount in the context menu, I get a message saying "unmount failed - device is busy".
Looking up with lsof what is causing the lock, it is always a nautilus process, which is CWD in some directory of the USB stick, e.g.:

nautilus 3927 henning cwd DIR 8,33 1024 3889 /media/usb1/2006-09-17

I am not sure, but it is possible, that this happens in conjunction with the problem described next - for example the above happened after I had to open the Terminal from nautilus in the way described below:


The next thing is, that the nautilus-open-.terminal function, since a while, nearly(but not always) any time I try to use it, it opens the terminal with a shell in my home directory, instead of the directory nautilus was navigatin in when I used the open terminal function.

What works is, right clicking on a directory icon, and using the "open with" dialog, and then chose "terminal" as an application to open the directory. That always works, but it's one menu level deeper to navigate each time, so I don't really like it.

Does the web outside know what's wrong?

Or should I really get up and ask the gnome-users mailing list??? :)

Fedora Core 8 VirtManager test

Another tool of which I tested a new version again: FC8 and VirtManager.

In general, VirtManager has gotten quite a bit better than the last version I tested (on FC6), but it still hase some annoyances to offer.

As nowadays(other than the last test), together with these annoyances come some very interesting features, especially managing qemu(+kqemu) and kvm (via qemu interface) instances in one single gui - now, my overall impression is, that it is a very interesting tool that can be recommended to fedora and GUI-Lovers, I think.

If you understand germen language, see my review/test here.

A deeper look at Google Ganeti

In the past few days, I got a deeper look in google ganeti, and asked many questions on their google group.

First, I got the impression that some things are not well working and lead to deadlocks and situations where you cannot start virtual instances anymore.

But it turned out to be all a matter of using the gnt-instance activate-disks command at the right place, using --force with the gnt-instance startup command, and --ignore-consistency when doing a failover...

Apart from a little bit ugly messages, Ganeti is really a nicely workoing management thing - even if I have not yet tested everything it can do for us.

My testing report in German is in our xen wiki.

Noch ein super Lieferant - cyberport

Herrlich, im Rahmen der "Bei Henning geht jede Hardware kaputt, und alle Dienste fallen aus"-Wochen, gibt es einen neuen Spitzenreiter:
Cyberport!

Dort habe ich am 22.9. einen 4 fach KVM Switch bestellt - und bis heute habe ich kein solches Gerät in einer Konfiguration, dass ich es richtig mit 4 Rechnern nutzen kann.
Sogar obwohl ich zwischendrin, da der Hersteller des eigentlich bestellten Gerätes, Dlink, keine funktionierenden Switches herstellen kann (viellecith etwas übertrieben, aber 20% Rückgabequote, die mir die Cyberport Mitarbeiterin berichtet hat, sind ja wohl sehr deutlich), nochmal tief in die Tasche gegriffen habe, und bereit war, insgesamt doppelt so viel Geld zu zahlen, wie die Dlink basierte Lösung kosten sollte!

Schönes Ding - Details hier.

Wahrscheinlich sind es nicht die Lieferanten, sonden ich, der momentan IT-Probleme magisch anzieht - ich sollte Schreiner werden!

Teresto Media - ein super-Provider

Auch Teresto Media muss, im Rahmen meiner diesjährigen "sämtliche eingekauften IT Dienstleistungen und Computerteile gehen kaputt"-Aktion besonders lobend erwähnt werden!

Für ein überragend gutes Problem-Management, und die Fähigkeit, das Neu-Aufsetzen einen IMAP-Dienstes nach einen Crash auf eine unmenschlich lange Zeitspanne hinauszuzögern.
Und für die Fähigkeit, auf Anfragen besonders freundlich und kompetentm, egal zu welchem Zeitpunkt immer mit der Antwort "es ist bald behoben" zu reagierern.

Da ich seit Donnerstag morgen keine Mails mehr lesen kann, die man mir an meine Hauptadresse geschickt hat, kann ich inzwischen guten Gewissens sagen: In der Zeit, die die brauchen, um das Problem zu beheben, wäre ich an jedes beliebige Datacenter der Welt zu Fuß gegangen, hätte dort aus dem Büro einen PC genommen, von Hand irgendein Linux installiert, und das imap Paket installiert, dass die Kunden wenigstens schnell wieder neu eingehende Mail lesen können.

Bei Teresto passiert nichts, außer, dass man mit jedem meiner Anrufe, die ich zwei Mal täglich mache, um mich nach dem Stand der Dinge zu erkundigen, immer unfreundlicher wird - mit mit einem Vorgesetzten oder der Geschäftsführung zu verbinden, wurde mir eben sogar von dem freundlichem Mitarbeiter, Herr Hoffeld, strikt verweigert.

Also blogge ich erst mal, um diese wertvolle Information über die Qualität der Leistungen der Firma auch allen verfügbar zu machen, die sich vor dem Abschließen eines Vertrages im Internet über die Firma informieren - und dann versuche ich es gleich mal mit einer Email an die Verantwortlichen in dem Unternehmen...

Kleiner Tip:
was Web- und Mailhosting im eher kleineren Bereich angeht, ist man sicherlich bei größeren Providern massiv besser und günstiger bedient - ich werde wohl zu Hosteurope gehen... wenn da alles kracht, habe ich wenigstens nur 1/4 des Preises gezahlt dafür.
Dass ich in Zukunft, bei welchem Provider auch immer, das Backup meiner IMAP Ordner selber erledige, ist auch logisch. Auch wen noch nicht klar ist, ob alte Mails noch da sind, oder weg - Herr Hoffeld sagt, alles sei noch da.
Dem kann ich aber nichts mehr glauben, da er auch seit Donnerstag Morgen sagt "heute Mittag ist der Server wieder in Betrieb - nur eine genaue Uhrzeit kann ich nicht nennen" - welches "heute" er meint, ist mir aber noch nicht ganz klar...

FAI Workshop

The weekend before this one (9th until 11th of November 2007), we had this year's officiual FAI developer workshop.

Even if some people were missing, we had a productive and interesting weekend!

See the details at the FAI Wiki.

As soon as I find time to get the photos from my camera, you will see how it looked :)

Ganeti - a Xen management tool that doesn't suck

Finally I found another tools for managing some aspects of a Xen infrastructrue that does not suck: Ganeti

It is small in code size, easily installed and does the things it promises quite easily.
When testing it, I found some bugs, most of which are more or less cosmetic - so, it feels a little bit more like a 0.5 version but a 1.2beta, but a quite good 0.5.
I saw quite some >1 version and >0.5 version tools for Xen which did suck quite a lot - no need to mention them again explicitly.

So, the way to have some nice set of tools to work nicely on a Xen infrastructure is made a step further. For now, my major favorites are:

• Xen-tools, which I combine with FAI for setting up a Domain and installing it
  
• Xen-shell, which I find very nice to give people access to manage a single Xen instance on a Node, without giving them acces to xm directly, while gicing them a similar power over a single instance
• Ganeti for getting setup and management of multi-node Xen clusters with a nicely working high availabiliy feature, that helps me do things for which I otherwise had to fiddle quite a lot with drbd on my own.

It should be mentioned explicitly that the developer group of ganeti is, same as Steve Kemp, the maker of xen-shell and -tools, a very cooperative group which takes feedback very seriously.

Keep Up The Good Work, guys!

OpenQRM vs. vendor lock in???

So, I thought I'd forget about the (not very successful)results I've got when investing some days into testing and evaluating - trying to build and install and get started with OpenQRM.

But then, I read this and that.

Really, guys, as for Vendor lock in, I think OpenQRM sits in the glass house:

My install is (because OpenQRM forces man own versions of libs available in each distribution, and not even the JDK is reused) 500MB - big binary blob.

It has been already discussed widely on their mailing list that OpenQRM wouldn't even let me change an IP address easily, because it "thinks" it should not be done, even if I _know_ exactly that the change cannot cause harm. So it enforces(or locks me into) a very specific structure and workflows to handle my datacenter.

When I decide to start using this as a main part of my infrastructure, and the whole things stops working, I guess I am totally locked in to buy support from qlusters, as it will take ages to understand the whole build system and source code before anything can be fixed.

Even for security updates of all the dependecies of OpenQRM I am locked in: instead of using the ones that come with a distribution, it brings own versions for many of them. If a grave security problem is found, I have to wait for the QRM developers to bring a security fix, instead of simply relying on the security support of my distribution vendor.

Maybe I am completely wrong and look at the things from a strange perspective, but that's my current opinion on the things.


To keep this post positive and constructive, some proposals for getting leaner, cleaner, and forcing less to your users:

1. Build against standard versions of libraries as they come with some distribution. If you have trouble building and developing against the multitude of lib-versions in many distributions, I'd say, better concentrate on one distribution and create a clean and small package. If your product is cool enough, there will be package maintainers from other distributions that will package it one day - or a customer will pay you to make it running on his favourite distribution. OpenQRM forces so many things, settling for a specific OS for the management server is a minor thing then. This has many advantages:
1. smaller code base at build as well as runtime leads to faster build, and easier error tracking
2. "free" security updates by distribution vendors
3. Enable OpenQRM being included in distributions and being available from their main software(yum/apt/yast...) repositories. They will not include it as long as it requires it's own combination of lib versions. other than the standard ones (imagine every app using it bringing it's version of rsync?!)
2. Instead of simply forbidding some things for all users(like chaning an IP address of a storage server), give experts some way to force doing a change you consider "not good", so people knowing what they do can force it.
3. In general, look for ways to keep your code base and runtime binaries small. By letting the user chose at install time to use a locally installed java runtime, for example, you can directly cut stuff in /opt for 100MB.
   

OpenQRM headaches

Gah!
I spent another 4 hours testing (or better: trying to get running!) OpenQRM. I already worked on getting it built, installed, and getting a Xen host and guest domains running and controlled with it for some more 24 hours over two days a while ago, without real success yet.

While OpenQRM seems like an interesting solution for people who want to build an control a complete datacenter in exactly the way the OpenQRM guys define it, it is definetely not a flexible solution, or something that can be "simply and easily" used for managing Xen systems - and that's what I tried to do with it.

Even changing the IP address of the storage server to be used wasn't easily possible - the way to go if you need that, is: adding a new storage server, and adding all the images that are available on this server again, manually, then changing all systems using this storage server to use the new one.

Matt, the project manager, is very helpful and responsive when I ask things on the meiling list, and he's a really nice guy. Still, I cannot give the tool good credits for the things I am testing it for: having a management console for a Xen server and infrastructure.

There are some other downsides of openqrm: it's packages come with a lot of stuff that is put into /opt - a lot of libraries which can be found on any system are doubled here, just in different versions.
I've been explained that is because "the other versions of these libraries don't fit our needs".
So I have busybox, rsync, php5(I don't even unsderstand what this is for, as the web interface is running as a servlet in tomcat), and others lying around in /opt in some random versions.

What about security updates to these? Will OpenQRM developers make a new release when any of these has security problem? Will I (easily) be able to install a new version of these libs myself, and OpenQRM will run fine with the security fixed version?
Will there be simple installable packages from the main software repositories of any distribution?
Probably none of these can be answered with a "yes".

These things are all good for the people offering support for the software - if any problem in any of these areas arises in your production datacenter after you decided to use OpenQRM, they will make quite some cash - if you have it. If not, you're lost.
I think this is vendor lock-in at it's best - exactly the contrary what you want to get if you decide to use Open Source software.

It's definitely a project that has it's niche where it will find it's users, but I don't see me becoming a big fan of it...
Still, I will look at it again in a while, to see if it made some progress. :)

Sunday Sunday

I really made it - not touching a computer for 24 hours! I must admit, even if I always fear these situations, because of all the stuff I miss to get done in that time, it's a good thing to do. Feeling much more relaxed, and more concentrated to work on my projects.

Today, I managed to get some measures against the spam in our Xen wiki done, and I finally made some first electronic design documents for the FAI GUI I started to work on.

Now back to some coding - a piece of software with a design, only, wil not be much of a use...

Rant on Ruby ambiguity and syntax woes

I've been part of the Ruby on Rails hype, as many others - I also wrote two small programs that are quite regularly used by me and Thomas, for whom I wrote a little helper to coordinate FAI development. Admittedly, neither of them ever made it being good enough to be published.

So, at the first glance, it looks very easy to do simple stuff with ruby/rails, and you can do some nice things very fast, without the hassles and complexities of setting up a Java project.

Still, the more I look at it, the more I wonder, if Ruby/Rails is a good way to go.
In developing a prototype for the new FAI GUI, I will check some interesting things:

• What happens if you start writing an app that doesn't use a database to store it's data, but, say, the filesystem?
• Will it be easy to use the model classes in an external lib, for creating other frontends than web-based ones? (most important: command line!)

And some other stuff...

But, I've come here to rant about some ruby stuff that starts to annoy me.

What about ambiguity?

I admit, it's saves some characters, but why do we have << class="blsp-spelling-corrected" id="SPELLING_ERROR_7">methods for the same functionality, and is it really necessary, to use special characters as method names? We learned from Perl that languages with too much ambiguity and shortcuts are likely to become "write only" - and not even the very authors of programs dare to touch them ever again, because they fail to understand them as soon as they stopped looking at them for more than 2 days. Don't get me wrong - I'm really impressed by the stuff Perl hackers can achieve with some lines of code - but the ambiguity and sometime cryptic syntax, implicit variables and stuff it provides stop me from learning it since ages. The more I get into Ruby, the more I fear Ruby will become the - fully object oriented - Perl of the next decade. Nothing to win here... In case you are tempted to tell me "this is just a single example" - yeah, right, what about:

File.open(path) do |file|
while line = file.gets
 p line
end
end

Two things here:

• What does the |file| syntax tell me as a human reader, exactly? (I searched, but never found a decent description of this
• Why do I need two blocks for one parsing action?
  

I use this syntax myself to parse a file, but the more I look at it, the less natural it looks to me.

In Python, the same looks like this:
f = open(path)
for line in f:
 print line

This is:

• Two lines less code
• only one loop block
• no special chars that are never really well explained (assume the : be a part of normal written human language)
  

back alive... new plans

Finally, we have finished our Xen Book, and finally, I have a bit recovered from the stress it defintely imposed to me - working full time a Java devloper, and being Beta-Tester for the most hyped(and not always most promise-keeping and working-as-expected) virtualization software on the planet at the same time is quite some thing to do.

The Xen developers give us a great software for free, but they intentionally let the distribution vendors do the job to build and package ist nicely and make their users the well working features available - as Simon Cosby put it in some private mail - "We're providing the engine, not the car."
They build their own car on top of it, as XenEnterprise.

Still, it was a great experience, and the next book is in the plannings - watch out for announcements about this.

For now, I want to get deeper and back to FAI development. Most important things I have in mind:

• Create a GUI to administer FAI
• Introduce a better hierarchy of the configspace
• Get more clean samples and docs of multi-distribution installs (e.g. Skolelinux, Windows, Fedora as the main candidates)
• Make fai-cd work with non-debian distributions

I won't get bored, I guess...