thx again for the info... more q's follow...
On Jun 13, 2004, at 10:27 AM, christophe varoqui wrote:
for right now ill probably use mdadm for the boxes we _need_ to be
supported, and ill start playing with dm on my debian ones. Just as a
side, is there a dm patch for 2.4.x that i could try?
the multipath tools rely heavily on sysfs and 2.6 hotplug. There is a
kernel patch for 2.4 on ftp://sources.redhat.com/dm/ but you won't get
far with that.
thats what it looks like. i imagine 2.6 is not till rhel4?
for right now it seems that ill just have each server in the cluster on
a different fabric, which should give me some measure of reliability.
is unloading/loading the module really the only way to rescan the bus
with the qlogic hbas? i hate having to take an outage to add a vdisk.
no, you got the following commands at disposal :
echo "scsi add-single-device 0 1 2 3">/proc/scsi/scsi
echo "scsi remove-single-device 0 1 2 3">/proc/scsi/scsi
where
0 = host adapter #
1 = bus #
2 = target #
3 = lun #
alternalively, you can google for a wrapper script written by Kurt
Garloff (I guess) named scsi-rescan.sh
sweet, thx. HP also seems to have their version of the same thing. ill
play with that for a bit.
could the RH kernel be built without REPORT_LUN? would that do it?
whats your take on using the RH, HP, or qlogic versions of the hba
driver? is one better than the others? how does the failover work? is
there a doc somewhere with the module params?
That is all a question of support.
For myself, I only use 2.6 for dev, which comes with the Qlogic
maintained driver.
where can i find info on the qlogic hba failover setup? ive heard
people talk about it but cant find any docs anywhere... what does it
do? how does it work?
thx again!
--
Guil Barros - UALR Computing Services - Networks Project Manager
gfbarros@xxxxxxxx - http://www.ualr.edu - 501-569-3345
"Because I dislike being quoted I lie almost constantly when talking
about my work." - Terry Gilliam