Provision Compute Engine instances

This page describes how to provision Compute Engine instances for Oracle Database deployment using Google Cloud NetApp Volumes.

To provision Compute Engine instances to host Oracle Database, complete the following tasks:

  1. Create the Compute Engine instances.
  2. Configure VPC firewall.
  3. Configure hostname, DNS, and /etc/hosts.
  4. Prepare the operating system on DB hosts only.
  5. Capture the IQN of each database host.

Create the Compute Engine instances

Create the following three Compute Engine instances in different zones of the same region (zonal failure isolation).

VM Zone Machine type Boot disk Network Purpose
oracdb1 us-west1-a n4-highmem-8 (sample) or c4-standard-* OL 10, 50 GB Hyperdisk Balanced (OS only) oracle-vpc or oracle-subnet, gVNIC Primary database
oracdb2 us-west1-b Same as primary OL 10, 50 GB Hyperdisk Balanced (OS only) Same Standby database
oradg-obs us-west1-c e2-medium OL 10, 20 GB Hyperdisk Balanced Same Fast-Start Failover Observer (Instant Client only)

For more information, see Create and start a Compute Engine instance.

While creating the instances, consider the following:

  • Network tier: Select Premium when latency or egress (>~200 GiB/month) matters and Standard for lower TCO in dev/test.
  • Security: Enable Secure Boot, vTPM, and Integrity monitoring on all three instances.

  • Storage: the boot disk holds only the operating system. /u01, Grid/DB homes, staging, and all ASM data use NetApp Volumes iSCSI volumes.

Configure VPC firewall

Configure VPC firewall to allowlist TCP port 1521 between all three instances for Oracle Net redo transport and Observer connectivity. Missing rules break Oracle Data Guard replication.

  1. Create VPC firewall rules with the following specifications:

    • Name: allow-oracle-net-dbhosts
    • Network: oracle-vpc
    • Direction of traffic: Ingress
    • Action on match: Allow
    • Sources: three /32 IPs, TCP 1521.

    You can mirror egress if required.

  2. To validate connectivity from each instance, run the following commands from each Compute Engine instance toward each peer IP:

    sudo dnf install -y nmap-ncat
for tgt in <oracdb1-ip> <oracdb2-ip> <oradg-obs-ip>; do
  nc -zv -w 5 "$tgt" 22
  nc -zv -w 5 "$tgt" 1521
done
    Port Expected output Meaning
    22 Connected SSH path works
    1521 Connection refused Firewall open; Grid listener starts in Step 6.1.5 (static DG entries in Step 8.3)
    Either Timeout Fix firewall or routing

Configure hostname, DNS, and /etc/hosts

On all three Compute Engine instances, configure the hostname and DNS resolution so that the forward and reverse name resolution works for Oracle Net, the Broker, and the Observer.

  1. Set the hostname and add /etc/hosts entries on all three instances.

    Substitute the Compute Engine instance internal IP addresses and run the following commands. You can retrieve the internal IP addresses from the Google Cloud console by going to the Compute Engine > VM instances list, and then looking at Internal IP column.

    # Run on each VM, substituting the local short name (oracdb1, oracdb2, oradg-obs)
sudo hostnamectl set-hostname HOSTNAME.example.internal
    # Run on every instance
sudo tee -a /etc/hosts >/dev/null <<EOF
# Oracle Data Guard peers + FSFO Observer
<oracdb1-ip>    oracdb1.example.internal    oracdb1
<oracdb2-ip>    oracdb2.example.internal    oracdb2
<oradg-obs-ip>  oradg-obs.example.internal  oradg-obs
EOF

  2. Validate name resolution from each instance.

    ping -c 1 oracdb1 && ping -c 1 oracdb2 && ping -c 1 oradg-obs

Prepare the operating system on DB hosts only

Prepare the OS on oracdb1 and oracdb2 for Oracle Database 26ai by installing the preinstall package, creating users and groups, installing iSCSI and multipath packages, and configuring the iSCSI initiator.

  1. Outbound HTTPS to yum.oracle.com (Cloud NAT or internal mirror on private subnets).

  2. Install the Oracle Preinstall package on both oracdb1 and oracdb2.

    sudo dnf install -y oracle-ai-database-preinstall-26ai \
  || sudo dnf install -y oracle-database-preinstall-26ai \
  || sudo dnf install -y oracle-database-preinstall-23ai

  3. Create the grid user, ASM groups, and add the oracle user to the ASM groups.

    sudo groupadd -g 54327 asmadmin; sudo groupadd -g 54328 asmdba; sudo groupadd -g 54329 asmoper
sudo useradd -u 54322 -g oinstall -G dba,oper,asmadmin,asmdba,asmoper grid
sudo passwd -l grid; sudo passwd -l oracle
sudo usermod -a -G asmdba,asmadmin oracle

  4. Install iSCSI, multipath, and JDK packages.

    sudo dnf install -y iscsi-initiator-utils device-mapper-multipath sg3_utils java-21-openjdk-headless libxcrypt-compat

  5. Verify THP and time synchronization.

    cat /sys/kernel/mm/transparent_hugepage/enabled   # expect [never]
timedatectl; chronyc tracking

  6. Configure SELinux, firewall, and iSCSI initiator settings, and then reboot.

    sudo setenforce 0
sudo sed -i 's/^SELINUX=.*/SELINUX=permissive/' /etc/selinux/config
sudo systemctl disable --now firewalld
    sudo cp -n /etc/iscsi/iscsid.conf /etc/iscsi/iscsid.conf.orig
sudo sed -i '/^[#[:space:]]*node\.session\.timeo\.replacement_timeout/d' /etc/iscsi/iscsid.conf
echo "node.session.timeo.replacement_timeout = 120" | sudo tee -a /etc/iscsi/iscsid.conf
sudo systemctl enable --now iscsid
    sudo reboot

Capture the IQN of each database host

After the reboot, capture the iSCSI initiator name (IQN) of each database host. You need these IQNs later to create the NetApp Volumes host groups.

  1. Copy the IQN from oracdb1 and note it somewhere safe.

    sudo cat /etc/iscsi/initiatorname.iscsi
# InitiatorName=iqn.1994-05.com.redhat:abc123def456

  2. Repeat the previous step for oracdb2, and note its IQN.

Use one host group per host so that a single host's reboot or IQN regeneration cannot affect another host's NetApp Volumes iSCSI volume visibility.

Cloned instances: If both hosts share the same IQN, regenerate on oracdb2:

  1. Stop iscsi
  2. Clear /var/lib/iscsi/nodes/*
  3. New InitiatorName in /etc/iscsi/initiatorname.iscsi
  4. Restart iscsi

What's next