Everything is a Freaking DNS problem

Yesterday evening was the speaker dinner it was in some Turkish
restaurant near the conference venue .. I was sitting with some of
the Xen people and Till "I kill the Rainforest" Kampetter , some
interresting stuff has been talked about..I bailed out "early
because of my talk today .. "

Keynote time, Olaf Kirch on the past and the future of the linux
kernel. In his Dear Santa letter we noticed NFSv5. What is NFS v5 ?
He goes on discussing the end of binary only modules, one day there
will be a "big bang" and we won't be allowing binary kernel modules
anymore, this won't be such a problem since mostly only 3D stuff
isn't relevant anyway ;).

Olaf's tells us that he knows that his week is ruined when on Monday
morning he gets a mail "Very Important Customer reports bad NFS
performance. Please fix immediately" It shouldn't be since the obvious and easy fix is probably vi /etc/hosts .

During the break Alan ran up to me and made some remark about my Everything is a Fscking DNS T-Shirt I should have traded it for a HAt :)
Alan is giving his regular "Why do you need Linux-HA", and We (the open source people) can afford HA since we haven't spent an arm and a leg to expensive software and hardware introduction It's all about "Blame" management. He then goes on describing the new features in LinuxHA v2, build-in resource monitoring, sophisticated dependency model where you can even define dependencies on remote machines or constraints that define resource you don't want to see together on the same machines, or resources that have to be on the same machine. Time dependent constraints such as only failback during such and such time, but failover when it is required. Resource management becomes more important, I should have a closer look at OCF, the Resources will have multiple states, started-as-slave, started as master, stopped, running etc.. About complexity: Isn't the v2 model significantly more complex than the v1 model ? LMB answers : you should see it as more consulting opportunities.

As everybody keept telling it as time to drop the 0. from the version number or DRBD , so DRBD made a hughe version step from 0.7 to 8. As a shared storage box is is a SPOF we obviously prefer a shared nothing environment, DRBD enables this for us. The most important new feature in DRBD: Shared disk semantics Allow both nodes RW access to mirrored storage. Usefull for shared disk file GFS, OCFS2, GPFS It seems that theperformance cutback for using DRBD is only 4% (according to Bonnie) , supporting up to 4TB -128Mb :) A commercial DRBD, DRBD+, version is now also available from Linbit. A DRBD version pre 8 is about to be released when Philip gets back from the conference.

Csync2 , a nice feature is adding triggers to the event of a file transfer, e.g an apache restart could be triggerd after an apache config file has been modified. It uses a database sqlite to store it's metadata which makes it a bit heavier than one might want the environment.

Csync2 looks to me (and Clifford) as a good tool to keep a simple cluster's config/data files in sync. Not as a way to manage your infrastructure or as a replacement for cfengine. Another difference is that csync2 uses a meshed architecture as opposed to a central repository in Cfengine.

Gerd Knorr gave a good overview of Xen which was very well attended even while he was scheduled next to Harald who usually atracts full
auditoria. Having seen Ian giving his Xen talk in Swansea already there wasn't that much new in it for me.

Then there was some speaker with stuff about how to deploy Xen in large environments. I already knew most of that stuff hmm.. Been there , done that .. gave the talk :)

Just before the boattrip to Social Event in the museum of work there was a talk about Freenigma which a lot of people missed since they
wanted to go back to the hotel first (including me). The social event location reminded me of the Linux.conf.au delegates session venue, it's a great concept to combine "technology"museum visits with a social event. Nils Magnus announced that LinuxTag will be getting more support from GUUG and moves to Wiesbaden next year, as Germany is totally in Football chaos during June it will also be early May for a change.

After my disaster roadtrip to LinuxTag I decided to fly to Hamburg
rather than getting stuck for 8hours on German Autobans, so in stead
of 6-7 hours driving I spent 1 hour in a plane , catching up on some
reading :)

Day 0 for me is actually Tutorial Day 2 , According to the usual suspects it seemed like the easiest way to
get network access is inside a Tutorial room... so I sneaked into
the Xen tutorial stuff. As I've already seen Ian's presentation at
the UKUUG and , thoe most interresting part I saw was the actual
Live migration of Virtual Machines , which leadto a file question :

"Q: How do you handle authentication when migration virtual machines ? A: Not at all"

Ever had problems with slow phone call setup, according to this Linux Journal article it off course is a fine DNS problem :)

I was in a Radisson hotel last week I experienced some strange behaviour on their access points. I booted up my laptop and the first thing I saw was an IP collision with my default preconfigured ip and the surrounding network. Off course there was no address already in use but upon booting I send out an arp request for my own ip address in order to figure out possible collisions. Somehow their environment sends a positive reply back. Strange :)

Next I tried to get an ip config via dhcp however failed to get an answer. Since their accesspoint replied to all my arp requests I figured out to set a random default gw. Upon realising that they transparently captured my outging traffic it took me a while till I got redirected to their authentication page.

Then I ran into my next problem, I could access hosts on ip basis but dns seemed to be completely disfunctional. I am running my own cachnig dns server on my laptop but somehow the answers that I got back were bogus. After configuring 2 random external dns server my luck changed. I assume they were doing some kind of dns dns spoofing but failing to do it correctly.

Yet another example of you ill configured services make life much more
difficult than it should be :)

Pieter pointed us to rpmstrap. It took me 5 minutes work to get an empty Xen FC4 instance with it.

Took ME 5 minutes , got no clue how long my computer has been working on building the actual image since I started it yesterday afternoon and didn't look back at it till late in the evening.
After deboostrap we now have tool that can build Fedora Core 2, Fedora Core 3, Fedora Core 4, Yellowdog 4, CentOS 3, CentOS 4, Mandriva and Scientific Linux systems. Whow that's about all of the most important rpm based distros except for the Suse versions.

When I visited San Diego a couple of years ago I spent some time in the San Diego Computer Museum , I just learned that it is currently closed and looking for a new location, It would really be a shame if this museum whold have to close it's doors permanently. When first visiting boston I had to learn that the computer museum that once existed there then had merged into another museum (the Mos Museum Of Science iirc) , however when visiting almost none of the expected machines could be found. Some museums such as the Melbourne Museum and the Powerhouse museum in Sydney still have some interresting machines on exhibit but apart from the Computer History Museum in Mountain View (which I've not visited yet) I haven't found any location that gives a good overview of computer history.

I can imagine lots of machines that are still functional and that could have an educational value hidden in some basements al over the planet (including mine). So If anyone around is interresting in setting up a new Computer Museum in Europe/Belgium send me a mail.. I'll probably be able to provide you with some interresting hardware !

Update: just found out about the PC Museum in Switzerland :)) got to visit that some day :)

Or in short, Gnome. yes that indeed is the full abbreviation, according to http://uncyclopedia.org/ :)
Other fun links on uncyclopedia are RMS , ESR , Windows and Microsoft...

One of our cats is mssing :(

Alan clearly pointed out what the problem is why people can't get along : Insufficient drugs, so might I suggest to organise the next Kernel summit in Amsterdam for a change ?

Marcus J Ranum of Gauntlet, TIS and NFR fame just published the six dumbest ideas in computer security. It's a must read for everyone ! Apart from his top 6 list Marcus also mentions some minor issues, such as Domain Naming System.

I think he missed the top dumbest idea ever, one that tons of vendors are trying to sell you, the idea that security can be bought in a box, as product. The idea that if you implement antivirus product such and firewall so you are safe. It's not the fact that those products only enumerate badness, it's the fact that vendors try to make people believe that these products will solve their problems.

No matter how you look at it , security can't be solved by a product, Security is an everlasting ongoing proces.