Good!
No problems with the patch, no significant problems with the tools.
(The only undesirable side effect being significant nerve tickle for
using alpha software on a production server.)
What I�m interested in: has anyone else had any major run-time-problems
with the software-raid (981215 or january edition)?
My configuration:
P133, 64 MB ram, no swap (a 486 would probably do it, too)
4 x ibm IDE 16.8GB drives, promise controller.
No DMA (if 2.0.36 would support it, I probably still wouldn�t use it).
machine usage: pure NFS fileserver
performance:
read: 5MB/s
write: 2.5 MB/s
obviously nothing to get excited about, but it�s what I expected (IDE,
I/O-polling). Still several times more I need (fileserver on 10MBit
network).
some minor problem I stumbled over:
- raidhotremove gives "device in use". I suppose it should remove a
drive from an array that is completely mounted. This is possibly solved
with the january release of the patch/tools. But I didn�t dare to
upgrade, and don�t need raidhotremove anyway.
something that could eventually be considered an issue:
- CPU usage during raid rebuild is high, i.e. the system feels somewhat
clunky (means: when holding a key for autorepeat, the cursor "jumps").
If some other process produces constant CPU usage, the raid rebuild load
goes down and is no more noticeable. I suppose this behaviour is due to
simultaneous IDE-polling of four drives (which is hard for any CPU).
However, it�s questionable if it�s worth to change that behaviour. I
don�t care about it, and as soon as the array is built, all is perfect.
