AIX 5.3 and Oracle RAC ASM diskmanagement



Hi,

We are using Oracle/RAC 10GR2 with ASM for disk management of diskspace
used by the database. ASM requires that the disks are NOT included in a
volume group and also uses the first block of the disk to store
essential information. RAC 10GR2 does not need HACMP anymore.
This setup imposes a serious problem for AIX systemmanagement:
1) It is difficult to distinguish disks on the AIX servers since they
are not part of a volume group.
2) The usage of the first block by Oracle implies that chdev may not be
used to create a PVID, since chdev overwrites the first block. If chdev
is being used on a ASM disk, Oracle can not read that disk, and the
database is lost.
3) Since ASM disks are writable by all servers in the cluster,
reserve_policy is switched off, the probability that a disk is by
accident used for a volume group is not unthinkable, with fatal
consequences for the database.
4) Changes in SAN configuration may imply that existing devices have to
be rearranged with rmdev. Rmdev removes disk devices and other devices
nodes in /dev having equal major/minor number. Using sybolic names for
hdisk devices in the Oracle database by creating devices nodes in /dev
is not a viable option.

In my opion these are some serious integration issues between Oracle
ASM and AIX. When my DBA collegues reported this to Oracle via a TAR
Oracle commented: "Works as designed, AIX system management should not
worry about PVID's and never give PVID to disks which are being used in
a database. Oracle will find out itself how disks should be used."
The way ASM handles AIX disks is not acceptable for our AIX system
managers, and we are looking for way to set things up in such a way so
that risks are reduced.

Question: Does anybody have any clues for solving this problem, are
there ways to reduce probability on mistakes (using ASM disk to make
volume group). How can we find out where disks are used.

Thanks in advance.

Erik

.



Relevant Pages

  • Re: Oracle Performance -- Possible Disk Bottleneck
    ... confirm this is a disk bottle neck. ... ton of information in the statspack report, so I'm not sure what else ... so I'm hoping someone here has experience with SANs and ORACLE to help ... this is a disk bottleneck and that giving the database more spindles ...
    (comp.databases.oracle.server)
  • Testing 10gR2 x86_64 on Linux (CentOS 4.2)
    ... Has anyone run ASM on top of RAID? ... of a disk, ... This problem doesn't have to do with Oracle, ... "If you don't eat your meat, ...
    (comp.databases.oracle.server)
  • Oracle and I/O performance.
    ... dba's are complaining about performance of importing a database export. ... another path to the SAN so that both volumes are on separate paths but I ... Oracle is running DirectI/O, asynchronous. ... An iostat on the two disks shows the bps and tps for the disk doing the ...
    (Tru64-UNIX-Managers)
  • Re: Oracle Database 10g R2 and RHEL4 Update 3
    ... ASM automatically rebalances and effectively stripes over all disks ... If a disk fails, the data ... ASMLIB, remove the failed disk from the OS, physically remove the failed ... adding storage to ASM is easy and managed entirely by Oracle. ...
    (comp.databases.oracle.server)
  • Re: SSA disk sharing
    ... FYI about online datafiles with Oracle ... > SSA Disk Concurrent Mode of Operation Interface ... The operating system has a concurrent mode interface to use this ...
    (AIX-L)