As you might have noticed my blogs were not updated since sometime during the second-to-last weekend of May. This appearantly was due to a fatal disk error which eventually pulled down our whole web server. We had experienced some weirdness before the whole server died to the effect that we suddenly couldn’t log into MT(Movable Type). I searched in vain for information on getting out of the infinite “invalid login” loop. I tried every trick in the book, but to no avail.
In the beginning of the week we found out that the problem was graver than we had expected as the disk error brought down the whole server. Frantic troubleshooting by the venerable Søren Rabbe (my dad) eventually led to a reinstall, fortunately we could save the pages which were soon established in their static form which stayed there until now. Now the web pages are stored on a different disk from the system, it has become part of our backup regimen, and we are seriously thinking about a small RAID solution.
With this new setup in mind we have made some other changes behind the scenes of these blogs, the now interface with MySQL which should prove more stable in the face of disk errors, and generally prevent some of the problems experienced with the Berkeley DB solution which is the default for MT. Furthermore, I have used the feature in MT which allows templates to be stored as flat files along with their placement in the database. This means that I have the files even if I cannot access the database, since the biggest hassle of getting this back on its feet was reconstructing the templates, this will be a huge timesaver in case of future problems.
After I get Susan’s and our Mothers blogs up again, I will spend some time on some thing I’ve had planned for the past few month. It’s just minor updates, a bit more color, some small features, but I’ll be sure to tell about them in time.
You may have noticed that there is no content yet. It is indeed so as all our old content is locked in a database which I cannot connect to. I have read about similar problems, they were able to extract the content from their old weblog using the Berkeley DB APIs for Python. The problems I have experienced aren’t as bad as theirs, using the fabulous tool MT-Medic I got access to the authors database, so there is life somewhere there. To salvage the pieces I will first try to use the MT APIs to gain access to as much data as I can, if that fails I will try direct access using the Berkeley DB APIs for Perl, in any case I guess I’ll learn something from this experience (I’ve always wanted to learn Perl).
For now I’ll rest on the laurels and prepare for my next update.
Tags: Site Matters