Cricalix.Net

The title of this post is actually slightly inaccurate - in the past 15 years, the number of hard drive failures my personal hardware has had (as opposed to hardware I’ve worked with) can be counted on one hand, two if my memory is a bit off.

This morning, I had reason to thank the gods of RAID - one of my drives died last night, and I didn’t lose a bit of data (hah, a pun!).

So, one RMA raised, and I have to hope that the other drive doesn’t decide to follow in the footsteps of its brother.  Perhaps I should get two more 500 GB drives and run RAID5 with hot-spare…

Oh, and an LTO drive.

It is a given that an upgrade normally breaks something.  Take, for instance, the upgrade of my laptop from Kubuntu 7.04 to Kubuntu 7.10.  7.04 didn’t play nice with my NVidia video card - it’s an older card, so the legacy driver was needed.  To get things working, I ended up removing files and editing hidden files.  7.10 broke this hack, but once I grokked that, and did it the Right Way, it all worked fine.

As another example, take Yosemite Backup.  Version 8.50-sp1 worked great on $work’s RedHat-based boxes, and with some hammering (libattr, libacl), worked great on the Gentoo boxes.  I had to write a script to act as the init script (due to Gentoo’s use of runscript), but otherwise life was groovy.

Today I upgraded the RedHat-based boxes to 8.50-sp2, along with the Windows boxes.  That part went fine, so I moved on to the Gentoo boxes, and found that things didn’t work fine.  To be precise, the boot binary for Yosemite was complaining “Error 59:  Service not found”. Mind boggling that.  So, run strace against the boot binary.  Wait, no strace.  Emerge strace, run strace against the boot binary.  Trawled through about 1000 lines of system calls, looking for a clue.

About an hour after I started looking, I found what looked to be harmless - the binary tried to call /etc/rc6.d/YTBackup.  This won’t work on a Gentoo box, it plain doesn’t exist.  A few symlinks later, and everything just worked.  Given that tidbit, and the fact that the binary kept looking in /proc/ for the wrong pid, I think that SP2 has become more RedHat oriented, breaking compatibility with other Linux distributions as it does so.  Time for a polite e-mail to Yosemite about that.

Got a hankering to do some cooking this afternoon, so cook I did.

250 g butter, equal granulated sugar - blend until approaching fluffy.

2 large eggs - blend into the sugar/butter mix one at a time until well mixed (no separation).

Some amount of flour - I did this purely by feel and texture.  Self raising.

A good dollop of Lyle’s golden syrup, mixed in thoroughly.

A teaspoon or so of salt.

A whack of grated, dessicated coconut.  No idea how much, again by feel / experience.

Oven cranked to 170 C or so, a large heaped spoon of mixture into each mold on the tray.

Baked for about 12 minutes.

Yummy.

With the web side of my mail and web migrated to a Xen VPS, I spent this weekend moving my mail services across.

SASL gave me grief, as it usually does.  I’m not sure whether it was the plain or md5 SASL extension that solved my problem in the end (I’m using MySQL as the backend via auxprop), but it finally started working after installing both of them.  Before that, I was getting a wonderful “warning: xsasl_cyrus_server_get_mechanism_list: no applicable SASL mechanisms” from Postfix, which then promptly declared that it was having nothing to do with me and went off in a huff.

With the SASL issue solved, everything else was a breeze - I just imported my old configuration for Postfix, amending it where things had changed (variable name wise).  I had to use SJ Mudd’s RPM package, as the default CentOS packages don’t know what MySQL is as a dictionary type, but beyond having to compile my own package, it was painless.  Courier compiled nicely into RPMs, and is happily chatting to the data source that ties together my valid accounts for SMTP and IMAP.  TLS is all integrated nicely, webmail is running, DNS is changed.

Bet something breaks!

KDE has supported the use of the Print Screen button as the hotkey for running KSnapshot for ages; to the point I can’t remember when it didn’t work. I used this feature heavily in Feisty, as it meant a single keystroke captured the screen, and brought up KSnapshot to deal with saving it (to wherever I wanted, including the network [go fish io_slave!]).

Gutsy broke this. Bug reports show it broken in Tribe 4 and Tribe 5, so I’m surprised it made it into the release version (which has even had updates released). This was an upgrade installation, so I’ll have to poke around with a fresh installation to see if it’s just an upgrade issue.

Note to self, when testing backup software that needs a gigabit link to get full performance, use the gigabit card for the network connection, not the 100 Mbit/s card.  I’ve been doing speed tests to test the multiplexing of Yosemite’s Backup, and it was capping out at 600 - 650 MB/minute, even with 8 servers as clients.  Two days into the testing, I realise that I forgot to move the appropriate IP to the gigabit fibre link.  Did that just now, and the backup server is pulling data in at 1.1 GB/minute or more.