Rebuilding an AIX LPAR for Oracle RAC with Pure Storage FlashArray

Rebuilding an AIX LPAR in an Oracle RAC environment isn't just an OS reinstall—it's a coordinated dance across Clusterware, ASM, storage multipathing, volume groups, and HMC. One wrong move corrupts OCR, breaks ASM discovery, or leaves you with an unbootable node.

This guide delivers a safe, repeatable, production-grade process for Oracle RAC on Pure Storage FlashArray (FCP, MPIO). We cover everything from clean Clusterware shutdown to Grid Infrastructure validation on the new LPAR.

Goal: Minimize risk, preserve RAC integrity, and keep a rollback path.

Environment Overview

Component	Details
OS	IBM AIX
Database	Oracle RAC
Storage	Pure Storage FlashArray (FCP, MPIO)
Disk Management	ASM
Backup	mksysb
High Availability	Mirrored Oracle VG + Alternate rootvg

1: Cluster Shutdown & System Backup

Stop Oracle Clusterware cleanly and back up the system before any storage changes.

# ./crsctl stop crs

# /usr/local/bin/backup/mksysb.sh

Why it matters:

• Prevents OCR/Voting disk corruption
• Creates a known-good rollback point
• Essential before destructive storage ops

mksysb Backup Script

This script validates rootvg, checks space, creates/verifies the image, and enforces retention.

#!/bin/ksh

########################################################################

# Script Name: mksysb.sh

# Purpose: Create filesystem-based mksysb backup for AIX

# Author: adminCtrlX

# Location: /usr/local/bin/backup

########################################################################

BACKUP_DIR="/backup/mksysb"

LOG_DIR="/var/log/mksysb"

RETENTION_DAYS=7

DATE=$(date +%Y%m%d_%H%M)

HOSTNAME=$(hostname)

MKSYSB_FILE="${BACKUP_DIR}/${HOSTNAME}_rootvg_${DATE}.mksysb"

LOG_FILE="${LOG_DIR}/mksysb_${DATE}.log"

mkdir -p ${BACKUP_DIR} ${LOG_DIR}

exec > >(tee -a ${LOG_FILE}) 2>&1

lsvg rootvg || exit 1

REQUIRED_MB=8192

AVAILABLE_MB=$(df -m ${BACKUP_DIR} | tail -1 | awk '{print $3}')

if [ ${AVAILABLE_MB} -lt ${REQUIRED_MB} ]; then

  echo "Insufficient space for mksysb"

  exit 1

fi

/usr/bin/mksysb -i -m -X ${MKSYSB_FILE} || exit 1

/usr/bin/lsmksysb -l ${MKSYSB_FILE} || exit 1

find ${BACKUP_DIR} -name "*.mksysb" -mtime +${RETENTION_DAYS} -exec rm -f {} \;

2: Pure Storage FCP Driver Installation

Mount the NIM image and install the MPIO driver for proper pathing.

# mount ibm-nim-001:/images/aix /mnt

# cd /mnt/purestorage

# installp -acXgd /mnt/PureStorage devices.fcp.disk.pure.flasharray.mpio.rte

# umount /mnt

# shutdown -Fr

Why it matters: Generic AIX paths lead to inconsistent failover and poor performance.

3: Service Startup & Disk Discovery

Post-reboot:

# startsrc -s collectd

# startsrc -s xntpd

# df -gt

Capture disk inventory:

# for i in `lspv | awk '{print $1}'`; do

    SIZE=`bootinfo -s $i`

    SERIAL=`lscfg -vl $i | grep Z1 | awk '{print $2 " " $3}'`

    echo $i $SIZE $SERIAL

done

Share WWPNs (e.g., from FCS0/FCS1) with storage team for LUN allocation.

4: Pure Storage MPIO Configuration

Set failover algorithm on Pure disks:

# for i in `lsdev -Cc disk | grep PURE | awk '{print $1}'`; do

    chdev -l $i -a algorithm=fail_over

    lsattr -Pl $i -a algorithm

    sleep 2

done

5: ASM Disk Renaming & Ownership

Rename and chown disks per ASM standards (e.g., ASMDATA_Disk, 

ASMOCR_Disk, ASMVOT_Disk):

# rendev -l <disk_name> -n <asm_disk_name>

# for i in `cat /tmp/asm-disk`; do

    chown grid:asmadmin $i

    lkdev -l $i -a

    ls -lrt $i

done

DB team then discovers and resyncs in ASM.

6: Oracle VG Operations (Source LPAR)

Clone rootvg, mirror Oracle VG, and wait for STALE PPs: 0.

# alt_disk_copy -B -d <disk_name>

# extendvg oraclevg <disk_name>

# mirrorvg -S -m oraclevg <disk_name>

# lsvg oraclevg

7: Create Target LPAR (HMC)

In HMC GUI/CLI:

• Launch Create Partition wizard.
• Map Virtual Ethernet (VIOS network, VLAN if needed).
• Map Virtual FC (NPIV): Pair client/server adapters, auto-generate WWPNs.
• Activate to Pending state.
• Extract WWPNs: lshwres -r virtualio --rsubtype fc -m <sys> --filter "lpar_names=<lpar>".

Share WWPNs with storage for zoning.

8: Migration Window Operations

Split/export Oracle/Data VG for safe migration:

# cp -p /etc/filesystems /etc/filesystems_Backup_$(date +'%d_%m_%Y')

# splitvg -y <new_vg> -i <old_vg>

# varyoffvg <vg_name>

# exportvg <vg_name>

9: Boot Target LPAR

SMS boot > Set rootvg disk > Configure networking (mktcpip, /etc/hosts).

10: Storage Cleanup (Target LPAR)

# for i in `lsdev -Cc disk | grep Defined | awk '{print $1}'`; do

    rmdev -Rdl $i

done

11: Oracle Filesystem Recovery

# importvg -y <vg_name> <disk_name>

# chlv -n <new_lv> <old_lv>

# chfs -m <new_fs> <old_fs>

# mount <fs_name>

12: Clusterware Startup & Validation

# cd /opt/oracle/grid/home/bin

# ./crsctl start crs

# ./crsctl check crs

# ./crsctl status crs

# bootlist -m normal <rootvg_disk>

Final Checklist:

# lspv

# df -gt

# ifconfig -a

# netstat -nr

# lssrc -s sshd

# lssrc -s rubrik_backup

Key Takeaways

• Pure Storage MPIO is mandatory.
• Standardize ASM disk names.
• Mirror Oracle VG for binaries.
• Use alternate rootvg for rollback/DR.
• Always clean shutdown/startup.

Pro Tip: Document serials in /tmp/asm-disk. Recovery becomes a checklist, not guesswork.