On Fri, 22 Oct 2004, Tim Starling wrote:
You should
implement some form of automatic checkup for empty disk space, or
you should consider buying more disks. Disk space is cheap.
Ariel has 6x73GB 15K SCA SCSI drives (4 RAID 1+0 (146GB), 2 RAID 1
(72GB)), and is out of drive bays. Disk space is not cheap.
ditto. I have boxes of 9.1G drives, a few dozen 18G drives, maybe a
dozen 73G'ers, and exactly 9 146G drives. As size increases, cost
exponentially increases. Let me look through the "inventory" and see
what I can donate -- provided you can actually use them; most will be
fibre channel drives.
Are there enough U's left for a drive shelf? (Like an IBM EXP15?)
We certainly should write a script to shut down the
master database
server safely before it runs out of disk space. Perhaps we should also
submit a feature request to MySQL AB asking them to prevent this situation.
I don't see any "good" solution to such an event. However, database writes
failing immediately would get people's attention :-)
Might I suggest using LVM. At that point, a re-sync would not require
killing all writes. Simply freeze a snapshot and start a new log. Duplicate
from the frozen snapshot and begin the sync with the now-recording log.
(this is the prescribed method of making a live backup of a SQL db on
other platforms -- wiki doesn't use solaris, oracle, or veritas :-))
--Ricky