SiGNOUT had a couple of amd64 machines with an Opteron 244 1800Mhz CPU and 3Gb RAM that are not in use at his work. He allowed me to install the Marcuscom Tinderbox on it so now I have another tinderbox for testing ports to be committed. Unfortunatly, 4.11 isn’t supported on amd64, so I still have to test those on my old 1300Mhz Celeron.
I used some time to make it possible to share the tinderbox with other people. So far I found that the best way to do that, is to create multiple ports trees and thus multiple builds so all users can have a ports tree owned by their own UID to enable them to patch the ports they want to test. There must be an easier way to do this, I just haven’t found it yet. And haven’t written the patches either :-)
Another nice feature would be to put the work directories on a memory backed file system. With the 3Gb memory in this box, there’s plenty of space to speed up the builds that way.

After the last couple of disk failures and subsequent problems with the RAID card in beastie (one of two servers that comprise ftp.freebsd.org), Scott Long sent me another RAID card by way of Mike Silbersack to EuroBSDCon. On Monday, I finally built up the courage and got beastie removed from the DNS round-robin so I could have some peace and quiet time in a service window. In theory, the new card should be able to read the configuration from the disks and take it from there, but somehow I’m a tiny bit paranoid and I don’t really trust that. So I had borrowed a bit on our new tape robot and flushed all the data to tape.

Yesterday, I took the machine down and exchanged the cards and rebuilt the RAID set. The card did read the configuration alright, but I removed it and built a new one. So far so good. Next, trying to boot the machine. All went well until it detected the second RAID set. The commands to the second card timed out and the machine couldn’t boot. Somehow, I got the machine booted after several tries and just let it sit there to let it sync the disks with the new config. Late afternoon, Scott came online and suggested to remove aacp from the kernel config as it might be causing problems and we don’t use it anyway. Beastie booted happily and now it was time to restore the data from tape. Writing to 8 disks in a RAID5 set obviously isn’t the easiest thing to do, so I only got about 4MB/sec out of it while we found out that the backup server only had a 100mbit card as we pushed 10MB/sec to it.

This morning the restore finally finished and after a final reboot, just to be on the safe side, everything is running again. Please welcome the new and improved beastie!

The grey weather today makes it a good day to have some quality time behind the computer screen. The first thing I decided to do today was upgrade my ports tinderbox to the latest release candidate (2.2.0rc5) of the MarcusCom tinderbox software. It was quite a smooth upgrade except for a minor issue that there was only an SQL script for PostgreSQL for the version I was running and not for MySQL, which is what I use, so I had to write one.

It’s been quite a while since I’ve really looked at MySQL code and I must admit that I’m amazed of how many new features they have added. Several years ago, I started using PostgreSQL and once you use things like foreign key constraints, you never look back. I’m glad to see that MySQL added those features as well.

It’s been quite a busy week with amongst others a meeting in our local UUG and the first “julefrokost” this weekend at the other side of the country. Right now, I’m just waiting for the bus to the airport on my way to this years EuroBSDCon in Basel, CH. Like last year, there will be a BOF on FreeBSD ports. If you have something to discuss or just are interested in hearing the discussion, see the wiki page on the BOF for more information. See you all there!