After spending much time fighting the documentation in
raid0145-19981215-2.0.36 and raidtools-0.90, and attempting every FAQ,
HOWTO,
and search engine I can muster. There still seems to be a lack of
documentation on how to use 0.90's hot-failover.
I have two RAID5 arrays with a mess of drives (>MAX_REAL), but I really want
to stress test the configuration (and trust the utilities and hotspares)
before trusting it fully. In 0.40, there was ckraid. Although sometimes
painfully slow, I endured it. Now it is in kernel space. How does a Good
Admin force a RAID check or resync?
Also, regarding ordering. Without the simple construct of cylinder, device,
lun, and partition/slice (as in the SVR4 world -> c0t0d0s0), this "variable"
/dev/sdXY drive identification is becoming quite irritating. If I add or
remove a tape/cdrom/JAZ/Zip device from my primary controller, those devices
appear AFTER sdb - but BEFORE the array, changing all array subdevice
ordering. (neccesitating a number of novel raidtab files)
This has caused my secondary RAID5 device to spontaneously decide the last
disk on my second controller to disappear. Fdisk sees it quite fine. I know
the disk and partition are there. The device links are correct, and the
major/minor access to the device is fine. However, 0.90 will not re-import
that disk into the second array even after removing the earlier scsi device
on the primary controller. What command do I use to fix this condition?
My solution: copy data off, rebuild the array, copy data back. There is no
documentation as to how to "invalidate" a known-bad disk that the internal
RAID rebuild will not use.
More importantly: there seems to be a general lack of documentation
regarding
any form of recovery with the 0.90 series. Is it an oversight on my part? Is
there something a well-intentioned admin can do to help?
- Ian C. Blenke <[EMAIL PROTECTED]> "Short .sigs save lives."
