PRACE Award 2011

PRACE Award 2011

This week I was in Hamburg to attend the International Supercomputing Conference (ISC’11), the main event in High Performance Computing in Europe, where our paper received the PRACE Award this year. I was together with Peter receiving the award, but it is the hard work of the whole team that made it possible. So, congratulations to all the authors!

The video is available from the University of Hamburg stream servers:  http://lecture2go.uni-hamburg.de/konferenzen/-/k/12298

Split OpenAFS volumes in version 1.4.x

I had to split the volumes on my server, in order to ease the management. In the OpenAFS 1.5.x branch, the volume server has a vos split command, which is not available for 1.4.x installations. Moving the dat arraound is not an option, since it does not preserves the ACL data. I came up with the following script:

It has certain drawbacks, like it’s not a move command but a copy, because it’s based in the AFS up command, so it requires you to have enough free space on the partition. It is also required to use the full path of the directory to split like /afs/cell/somefolder/dirtosplit. Besides that, it helped me to automate the splitting of several huge volumes.

You can check the help with the -h option, and you can test that all variables are properly set before executing using the -t option.

Enjoy!

Continue reading

Fix the device path for a SCSI device with multiple HBA adapters

I was installing Bacula on my server, and I got the following problem: Because I have several SCSI adapters, something changed the enumeration order of the adapters, causing the SCSI devices to be enumerated in a different order. Now, for hard drives it is mostly a non-issue, as the uuid and LVM take care of mounting the correct partition on the right place. The exception to this is during boot, but that can be solved one way or another.

However, the situation is different for tape drives (and my autochanger). The change in order moved the generic SCSI device number, and swapped the number of both my tape drives. So it ended up in Bacula that the autochanger was not the autochanger, the tape drive of the autoloader was not the one in the changer, and the comands where sent to all the wrong devices. Besides changing the reference, I wanted to avoid this from happening again.
Continue reading

Hacking the Adobe Lightroom Catalog

I had a problem the other day.. I wanted to moved all my Photographs from my workstation to my server, and access them over the network. The main issue is that I keep all of them organized in Adobe Lightroom catalogs (v. 1.4.1), and it does not allow the catalogs to be in network drives. That’s not a big deal, as I prefer to have the catalogs local (just a couple of GBs) and the bulk of the files in the server (around 130 GBs).

Since I moved the files using a USB drive, the Catalog had all files referencing the location of the drive. So I first tried to “find missing folders” option to relocate them. Unfortunately it asks me to apply the changes to every folder manually, and that was too much work. Besides, it was renaming the folder entry to “/new-path/date”, which looks very ugly and renders the folder browser in the library useless.
Continue reading

The Matrix and the Three Laws of Robotics

This is just a mental excercise, so let’s play a little, shall we? =)

Mixing two Science Fiction worlds is always dangerous. However, both the Three Laws of Robotics from Isaac Asimov and the Matrix Universe from the Wachowski brothers explore the issue of machines becoming intelligent and having the capacity to destroy their makers, that is, us.
Continue reading

A Scanner Darkly

I was looking at this movie again last week, it’s really visually very attractive. Not that the story is not good, it’s also great (a Philip Dick story after all). I took afterwards some time to research on the movie, specially the difficulties they had during post production. Every frame was processed by hand by a huge team of animators doing digital rotoscoping all day… for 18 months!

I really like the result.