Hi,
Our outsourced developer has given up on using /dev/mapper/...
to set up Oracle RAC ASM as they ran into problems with one
RAC server (on RHEL 4.5) could see/access the ASM partitions
while the other 2 RAC servers can't. So we won't pursue this
option now.
They've managed (& are happy) to get Oracle 10gr2 ASMLib working
on all 3 servers across reboots using Linux device files, specifically
they've used the raw /dev/sdc1, /dev/sdd1, /dev/sde1, /dev/sdf1
for ASM partitions while /dev/sdg1 is mounted on a mount point as
ocfs2 partition (this ocfs2 serve as sort of 'quorum disk' for Oracle
RAC cluster, I think; I'm no DBA, just Unix admin)
Developer is now happy as data could be retrieved as
ASM partitions are stable across Linux reboots. However,
NetApp SAN admin is now unhappy as occasionally we're
getting "Fcp partner path misconfigured" alerts from the
Netapp SAN filer.
We suspect this alert is triggered by a change in the
mapping between /dev/sd.. & the actual disk chunk each
time Linux is rebooted (as shown by "oracleasm querydisk")
For eg, "multipath -ll" would show /dev/sdq1, /dev/sdw1,
/dev/sde1 & /dev/sdk1 are all pointing to the same chunk
of disk storage. Developer uses /dev/sdq1 initially but
after reboot, the same ASM partition would get remapped
to /dev/sde1 :
"multipath -ll" output (taking one of the ASM partition as sample):
mpath2 (360a980005672443963344933
706f536c)
[size=40 GB][features="1 queue_if_no_path"][hwhandl
er="0"]
\_ round-robin 0 [active]
\_ 8:0:2:3 sdq 65:0 [active]
\_ 8:0:3:3 sdw 65:96 [active]
\_ round-robin 0 [enabled]
\_ 8:0:0:3 sde 8:64 [active]
\_ 8:0:1:3 sdk 8:160 [active]
To further illustrate, say, before reboot :
[root@BMW11 ~]# /etc/init.d/oracleasm querydisk land1
Disk "LAND1" is a valid ASM disk on device [65, 1] <== correspond to say /dev/sdq1
After reboot,
[root@BMW11 ~]# /etc/init.d/oracleasm querydisk land1
Disk "LAND1" is a valid ASM disk on device [different_number, 1] <== remapped to /dev/sde1
The above "disk chunk" numbers can be seen in /proc/partitions
We're not certain if this change across reboots is the cause of
"partner path misconfigured", only a suspicion.
To verify our suspicion, I need someone to tell me the following :
in RHEL, there's something called /etc/udev/permissions.d/
50-udev.permissions.
Can someone give me a specific example configuration (with
actual code) of how I can make it such that only the device
file that I need (/dev/sdq & /dev/sdq1) are created whenever
Linux is rebooted
We noticed that for the ocfs2 partition (which is mounted to a
mount point in Linux), it consistently gets the same "disk chunk"
number but not the raw ASM partitions. I have no inkling why.
Or anyone has any startup script that would delete away all
the "undesired" device files just before Oracle ASM gets started
during a reboot - correct me if I'm wrong, if Oracle ASM can't
find one device file, it will scan for the next available device file,
is this correct?
Can I boldly request for someone to read the attached document
from Oracle, page 5. I'm not too good with this. The attached
document has some clue which I probably can't make sense of
it as I needed exact codes.
If you have a suggestion of where we have gone wrong with
/dev/mapper, appreciate it too.
Start Free Trial
