Skip to main content
NetApp database solutions

Provision Google Cloud NetApp Volumes iSCSI storage for Oracle Database 26ai

Contributors netapp-jsnyder
PDFs

Provision Google Cloud NetApp Volumes iSCSI block storage for Oracle Database 26ai high availability on Google Compute Engine. This procedure covers creating GCNV Flex Unified storage pools, defining host groups, creating iSCSI volumes for each database host, configuring Linux iSCSI and multipath, partitioning ASM backing devices, and mounting the /u01 filesystem.

Step 1: Create GCNV iSCSI pools

Create two Flex Unified storage pools, one in each database zone, to provide iSCSI volumes for the primary and standby hosts. Each database host uses volumes from its local zone's pool.

  1. Create two storage pools using the Cloud Console. Use the specifications in the table below and repeat the creation process for each zone:

    Pool name Zone Used by

    oracle-pool-a

    us-west1-a

    oracdb1 (primary)

    oracle-pool-b

    us-west1-b

    oracdb2 (standby)

    NetApp Volumes → Storage pools → Create for each pool:

    • Service level: Flex (not Premium)

    • Type: Unified

    • Zone: match the database VM zone (us-west1-a / us-west1-b)

    • PSA: connected to oracle-vpc

    • Capacity: sized for workload; use custom provisioned throughput/IOPS when redo, backup, or restore exceeds default headroom (up to 5120 MiB/s or 160K IOPS per pool, per product limits)

  2. Wait for both pools to reach READY status before proceeding. Scale pool sizes to your database footprint (the sizes in Step 3: Create the GCNV iSCSI volumes are examples):

    Note

    Default-mode (this guide): Flex Unified pools use Default-mode (--mode=default). Create pools and iSCSI volumes with Cloud Console or gcloud netapp. Volume replication, snapshots, and clones use Google Cloud APIs (Step 3: GCNV standby initialization).

Step 2: Create host groups

Create one host group per database host so each VM sees only its own volumes. The primary and standby hosts must not share GCNV iSCSI volumes to maintain independent storage.

  1. Create the host group for oracdb1 using the Cloud Console:

    NetApp Volumes → Host groups → Create

    • Name: oracdb1-hg

    • Region: us-west1

    • Type: iSCSI initiator

    • OS type: Linux

    • Hosts: paste the IQN from oracdb1 (the value of /etc/iscsi/initiatorname.iscsi)

    • Description: "Oracle primary host oracdb1"

    • Create

  2. Repeat the process for oracdb2 with the name oracdb2-hg and `oracdb2's IQN. The Observer host requires no GCNV resources.

Step 3: Create GCNV iSCSI volumes

Create five GCNV iSCSI volumes for each database host: one for /u01 and four for ASM backing devices. Each host's volumes must be created in its local zone's storage pool with its corresponding host group.

  1. Create the five volumes for oracdb1 in oracle-pool-a with host group oracdb1-hg. Use the specifications in the table below:

    GCNV iSCSI volume Size Use Multipath alias

    ora_<host>_u01

    100 GiB

    /u01 GCNV iSCSI volume — Grid/Oracle homes, staging

    /dev/mapper/ora_<host>_u01

    ora_<host>_data_01

    50 GiB

    ASM +DATA

    /dev/mapper/ora_<host>_data_01

    ora_<host>_data_02

    50 GiB

    ASM +DATA (striped)

    /dev/mapper/ora_<host>_data_02

    ora_<host>_arch_01

    100 GiB

    ASM +RECO

    /dev/mapper/ora_<host>_arch_01

    ora_<host>_fra_01

    100 GiB

    ASM +FRA

    /dev/mapper/ora_<host>_fra_01

    Volume names: letters, numbers, underscores only (no hyphens).

    Note

    Minimal layout (validation only): Two LUNs per host (*_data, *_reco) with arch_01p1→+RECO and arch_01p2→+FRA is acceptable for lab; production uses five volumes per Step 3: Create the GCNV iSCSI volumes.

  2. Create the five volumes for oracdb2 in oracle-pool-b with host group oracdb2-hg using the same specifications. For each pool, use NetApp Volumes → Volumes → Create — iSCSI, correct pool and host group, Linux. Record the following information:

Step 4: Configure iSCSI and multipath

Configure iSCSI and device-mapper-multipath on each database host to access the GCNV volumes through both storage portal IPs. Run these steps on oracdb1 using the primary pool's portal IPs, then repeat on oracdb2 using the standby pool's portal IPs. If host egress is restricted, allow TCP/3260 from each database VM to its GCNV iSCSI portal IPs (in addition to inter-VM TCP/1521 from Step 2: VPC firewall — allowlist TCP/1521 across all three zones).

  1. Discover targets, log in, and persist node startup:

    sudo iscsiadm --mode discovery --op update --type sendtargets --portal <ISCSI_PORTAL_1>
sudo iscsiadm --mode discovery --op update --type sendtargets --portal <ISCSI_PORTAL_2>
sudo iscsiadm --mode node --op update --name node.startup --value automatic
sudo iscsiadm --mode node -l all
sudo systemctl enable --now iscsid iscsi multipathd
sudo iscsiadm --mode session         # expect 10 sessions (5 GCNV iSCSI volumes × 2 portals)
sudo lsblk -o NAME,SIZE,WWN,VENDOR,MODEL

    After reboot, re-check before starting Oracle:

    sudo iscsiadm --mode session
sudo multipath -ll

  2. Configure device-mapper-multipath with defaults and blacklist rules:

    sudo tee /etc/multipath.conf >/dev/null <<'EOF'
defaults {
    find_multipaths     yes
    user_friendly_names yes
}
blacklist {
    devnode "^(ram|raw|loop|fd|md|dm-|sr|scd|st)[0-9]*"
    devnode "^hd[a-z]"
    devnode "^cciss.*"
}
EOF

sudo systemctl enable --now multipathd
sudo multipath -ll

  3. Add host-discovered WWID aliases to /etc/multipath.conf (do not guess — multipath.conf does not expand shell variables). Discover WWIDs:

    sudo multipath -ll
for dev in /dev/sd*; do
  [ -b "$dev" ] || continue
  printf '%s: ' "$dev"
  sudo /usr/lib/udev/scsi_id --whitelisted --device="$dev" 2>/dev/null || true
  echo
done

    Append concrete aliases for that host to /etc/multipath.conf, then sudo systemctl restart multipathd.

    On oracdb1, append:

    multipaths {
    multipath { wwid <host-discovered-wwid-for-u01>      alias ora_oracdb1_u01     }
    multipath { wwid <host-discovered-wwid-for-data-01>  alias ora_oracdb1_data_01 }
    multipath { wwid <host-discovered-wwid-for-data-02>  alias ora_oracdb1_data_02 }
    multipath { wwid <host-discovered-wwid-for-arch-01>  alias ora_oracdb1_arch_01 }
    multipath { wwid <host-discovered-wwid-for-fra-01>   alias ora_oracdb1_fra_01  }
}

    On oracdb2, use the same pattern with ora_oracdb2_* aliases, then:

    sudo systemctl restart multipathd
ls -l /dev/mapper/ora_$(hostname -s)_*

Step 5: Partition ASM devices

Partition the four ASM backing devices (excluding u01) with one GPT partition each for ASM consumption, then configure udev rules for grid ownership. Run these steps on each database host.

  1. Partition the four ASM backing devices with GPT and verify the partitions:

    HOST=$(hostname -s)        # oracdb1 on the primary, oracdb2 on the standby
for dev in /dev/mapper/ora_${HOST}_data_01 \
           /dev/mapper/ora_${HOST}_data_02 \
           /dev/mapper/ora_${HOST}_arch_01 \
           /dev/mapper/ora_${HOST}_fra_01; do
  sudo parted -s "$dev" mklabel gpt
  sudo parted -s "$dev" mkpart primary 0% 100%
done
sudo partprobe
sudo systemctl reload multipathd
ls /dev/mapper/ora_${HOST}_*p1     # expect 4 partitions

  2. Configure udev rules to assign grid ownership and trigger the changes:

    HOST=$(hostname -s)
sudo tee /etc/udev/rules.d/99-oracle-asm.rules >/dev/null <<'EOF'
KERNEL=="dm-*", ENV{DM_UUID}=="part?-mpath-*", ENV{DM_NAME}=="ora_oracdb*_*p?", \
    OWNER="grid", GROUP="asmadmin", MODE="0660"
EOF

sudo udevadm control --reload-rules
for part in /dev/mapper/ora_${HOST}_*p1; do
  dm=$(readlink -f "$part" | xargs basename)
  sudo udevadm trigger --action=change --name-match="/dev/${dm}"
done
sudo udevadm settle
ls -lL /dev/mapper/ora_${HOST}_*p1    # grid:asmadmin 0660

Step 6: Format and mount /u01

Format the ora_<host>_u01 GCNV volume with XFS and mount it persistently using UUID in /etc/fstab. The /u01 filesystem holds Grid home, Oracle home, and staging files.

  1. Format the multipath device with XFS and capture its UUID:

    HOST=$(hostname -s)
U01_DEV=/dev/mapper/ora_${HOST}_u01
ls -l "$U01_DEV"

sudo mkfs.xfs -f "$U01_DEV"
U01_UUID=$(sudo blkid -s UUID -o value "$U01_DEV")

  2. Add the UUID-based mount entry to /etc/fstab and mount the filesystem:

    sudo mkdir -p /u01
echo "UUID=${U01_UUID} /u01 xfs defaults,_netdev,nofail,x-systemd.requires=iscsi.service,x-systemd.requires=multipathd.service,x-systemd.after=iscsi.service,x-systemd.after=multipathd.service 0 0" | sudo tee -a /etc/fstab
sudo mount -a

  3. Create the directory structure with proper ownership for Grid and Oracle software:

    sudo mkdir -p /u01/app/oraInventory /u01/app/26ai/grid /u01/app/grid \
  /u01/app/oracle/product/26ai/db_1 /u01/stage
sudo chown -R grid:oinstall /u01/app/oraInventory /u01/app/26ai /u01/app/grid
sudo chown -R oracle:oinstall /u01/app/oracle /u01/stage
sudo chmod -R 775 /u01/app /u01/stage

    Reboot once and confirm /u01 mounts before installing Oracle software.

What's next?

To install Oracle Grid Infrastructure and Database binaries on prepared hosts, go to Install the Oracle Grid Infrastructure and Oracle Database software on both hosts.