Well, I've gotten a variety of responses to my previous query: Most of them
were positive. Someone under the address ADSML AT MAILANDNEWS DOT COM was
emphatically against raw volumes for any purpose. I'm interested in why; As I
look back over the list, it seems that there was some time ago (ca '95) some
question about recovering raw lv's, but that was described (by the person
expressing the reservations) as probably-unnecessary residual paranoia five
years ago.
I'm interested in any updated comments or experiences that folks have had
comparing raw LVs to JFS volumes.
Second: I'm considering reorganizing the locations of my DB and log volumes,
now that I have this cornucopia of disk. So, from first principles:
If you've got a dozen spindles, and you need to divide DB, log, and stgpool
space amongst them, what's the best organization?
I can see two endpoints:
1) the (mirrored) db volumes on separate spindles from the (mirrored) log
volumes, also separate from the stgpool volumes.
2) a chunk of each on each: each spindle has some recovery-log, some DB, and
some storage space.
2) is a miserable solution: contention city.
1) would be fantastic, if my recovery-log needs were anywhere close to a
spindle size; they're far less.
Currently I envision
four spindles devoted to DB duty,
two carrying log volumes and some very light usage stgpool space
the rest doing normal stgpool duty.
I welcome comments, criticism, whatever.
Allen S. Rout
|