It’s great that these things keep getting easier. I used to have to tweak .gnomerc and .uim files and running uim-xim at startup… Probably that’s a thing of the past. Here’s how you do it the modern way.
yum install scim-lang-japanese
Next, go to System | Preferences | Input Method. This brings up the IM Chooser dialog. I selected IBus because it was marked recommended. Then in Input Method Preferences, and choose the Input Methods tab and select the input methods you want. That’s it! You can swith to Japanese input with Ctrl – Space.
My hard disk started showing signs of slow and painful death, so after I backed up all my data I confirmed with smartctl that yes, it’s dying.
smartctl -t long /dev/sda smartctl -a /dev/sda
This looked like a good opportunity to try something new, at least Kubuntu instead of Ubuntu. So I went for the much celebrated Fedora 11.
The live CD
My half-broken hard disk was still in the computer when I started the Live CD and Fedora scored some plus point for showing a warning about the disk, and a very cool GUI for the smartctl self tests. Next I replaced the hard disk with two new ones, and started the Live CD again and wanted to use the fancy GUI for smartctl. I looked through the menus, the Gnome applets but couldn’t find it. Wonder where it is, it’s a shame to hide this superb tool in a dusty corner. So I did the smartctl the old way. (By now I know where it is: Applications / System Tools / Palimpsest Disk Utility.) Then I had some MAJOR problems with the disk partitioner. I have two disks of the same size and I used to have them in a software RAID1, but that’s not what I intended for this time. The installer’s partitioner oddly showed me only one disk, and with an odd name. It looked like some kind of mirroring setup, so I checked
lvs but I couldn’t find anything. For hours I didn’t know what hit me, when finally I figured out it’s
I was very disappointed by this
dmraid thing. Why does the installer try to be so damn smart and do something I don’t want without asking me? And then why is it so damn hard to turn it off? I passed
nodmraidkernel param to the installer as suggested by google sensei, but no dice. Finally I managed to turn the bloody thing off with:
- Rest both partition tables
- Start the installer, and if you’re lucky it will ask you to reinitialize the partitions, go for it and exit the installer
dmraid -an and
dmraid -x, or something like that, I don’t remember exactly
- Just to be safe, I created my partitions in
cfdisk before going back to the installer.
Finally the installer showed the disks as they really were and I could get on with the installation.
So far so good. I liked that upon first start-up the system offered to setup ntp synchronization. It’s also nice that the nfs client tools work well out of the box, in some systems I often need to tweak
/etc/init.d scripts to get nfs partitions automatically mounted at boot. And I could setup my RAID1 partitions just fine in my good old-fashioned mdadm way.
Boy, is this complicated? Yeah, but hey, replication is a serious business…
Anyway, I just managed to set it up, and the one thing that wasted most of my time was broken stored procedures. The lesson here is, before publishing a stored procedure for replication, recreate it at the source first! (Right click, Modify, F5)
Actually replication might be overkill. Instead of that, daily full restore from daily full dumps might be a viable alternative.
Dump database T-SQL hint:
BACKUP DATABASE [SameName]
TO DISK = N’D:the long long pathSameName_backup_200906091739.bak’
WITH NOFORMAT, NOINIT,
NAME = N’SameName_backup_20090609173911′,
SKIP, REWIND, NOUNLOAD, STATS = 10
Restore database T-SQL hint:
RESTORE DATABASE [SameName_Restore]
FROM DISK = N’D:the long long path200906091105.bak’
WITH FILE = 1,
TO N’D:the long long pathSameName_Restore.mdf’,
TO N’D:the long long pathSameName_Restore_1.ldf’,
NOUNLOAD, STATS = 10