Oracle RAC 19c installation (Grid Infrastructure + ASM + RAC database)

Installs Oracle Grid Infrastructure 19c (Flex ASM, SCAN) and an Oracle RAC 19c database from scratch on two or more nodes: network and DNS plan, shared disk preparation (udev or ASMLib; multipathed or single-path), users and SSH, cluvfy, silent gridSetup/runInstaller with the RU applied during installation, a RAC CDB with dbca, an application service and failover tests. The result is a RAC database with an open instance on every node and a service reached through SCAN. For Grid + ASM on a single server use the Oracle Restart guide; for day-to-day operation of an installed cluster (adding nodes, OCR backup, eviction) open the RAC operations guide.

Topology: 2–8 nodes + shared disk (SAN/iSCSI) + SCAN in DNS

Before you start

• All nodes run the same supported operating system release and the same kernel. …On OL9/RHEL9, 19c can be installed from the 19.3 installation media only by applying the current RU and the one-offs from the Release Notes during installation with -applyRU / -applyOneOffs (at the time of writing the Release Notes recommend 19.32 or later; also check the minimum OPatch version in the RU README).…OL/RHEL 7.9: 19c Grid Infrastructure and RAC are supported; the base 19.3 installer recognizes OL7, and the RU is still applied during installation with -applyRU. Minimum kernels: UEK4 4.1.12-124.19.2, UEK5 4.14.35-1818.1.6, UEK6 5.4.17-2011.4.4 or RHCK 3.10.0-862.11.6.…On OL8/RHEL8 the 19.3 base installer does not recognize the operating system; a current GI RU is applied during installation with -applyRU.…
• At least 8 GB RAM on each node (much more for production), swap according to RAM (4–16 GB RAM → equal to RAM; above 16 GB → 16 GB), at least 1 GB in /tmp, at least 100 GB free for /u01 (Grid home ~12 GB, DB home ~10 GB, plus room for the RU and logs).
• On node 1, LINUX.X64_193000_grid_home.zip, LINUX.X64_193000_db_home.zip (eDelivery / oracle.com), p6880880_190000_Linux-x86-64.zip (latest OPatch, MOS patch 6880880) are ready under …, together with the extracted GI RU directory …. A My Oracle Support account and a license are required.
• Each node has two network interfaces with the same names on all nodes: public … (…), private … (…). The private network is dedicated to the interconnect, non-routed, at least 1 Gbit (10 Gbit recommended); the MTU is the same on all nodes.
• The IP plan is ready and the VIP addresses are free (not yet used by any server): …… public … / VIP … (…-vip) / private … (…-priv)…; …….
• In DNS, the name … resolves round-robin to 3 different IP addresses on the public network (A records); reverse (PTR) records exist. The SCAN name and its IPs are not added to /etc/hosts (only one IP is read from the hosts file, so SCAN load distribution does not work). Defining the node names and VIPs in DNS is also recommended; this guide additionally writes them to /etc/hosts.
• Time synchronization on all nodes is done by chrony from the same NTP source; the clock offset is below 1 second. (When chrony is running, Oracle's CTSS service stays in observer mode.)
• The shared disks (SAN/iSCSI/FC) are visible on all nodes, the same size, with no file system, LVM or old ASM header on them. For the CRS disk group …at least 3 disks (each at least 10 GB)…at least 1 disk (10 GB or more recommended)…; DATA and FRA according to data size. …Each LUN has at least 2 paths on all nodes; in this guide the multipath aliases are written identically on all nodes and the udev rule matches on ENV{DM_NAME}.…The disks are single-path; the udev rule matches on ENV{ID_SERIAL} (on a multipathed SAN choose 'multipath: yes').… Finding and comparing WWIDs: Storage preparation.
• The nodes can reach the operating system repositories (…Oracle Linux: ol7_latest; RHEL 7: subscription (Server + Optional, ELS)…Oracle Linux: ol…_baseos_latest / ol…_appstream… / ol…_addons…; RHEL: subscription…) or an internal mirror.
• For ASMLib v3, oracleasmlib-3.x.el….x86_64.rpm (oracle.com → Oracle Linux ASMLib download page) is under … on every node.
• You have root access on all nodes; a maintenance window, server/SAN snapshots and a fallback plan are ready. The firewall (the interconnect must be fully open) and SELinux decisions have been agreed with the security team.

Phases of the guide

• Operating system preparation (all nodes) — Set the hostname (…) · Update the system and install basic tools · Install the Oracle preinstall RPM · Create the ASM groups and the grid user · Set passwords for the grid and oracle users · Set the limits for the grid user …
• Preparing the shared disks — Install and enable DM-Multipath · Write /etc/multipath.conf · Load multipath and verify the paths (…) · Write the udev rules · Apply and verify the udev rules · Install the ASMLib software …
• Software, SSH equivalence and pre-installation checks — Set ownership of the installation files · Extract the Grid software into the Grid home · Update OPatch in the Grid home · Install the cvuqdisk RPM (node 1) · Set up SSH equivalence for the grid user · Copy the cvuqdisk RPM to the other nodes …
• Grid Infrastructure installation — Write the Grid response file · Install Grid Infrastructure (gridSetup, with the RU) · Run orainstRoot.sh (…) · Run root.sh — … (order …) · Run the configuration tools · Verify the cluster status …
• Creating the ASM disk groups (DATA, FRA) — Create the DATA and FRA disk groups · Mount the disk groups on all nodes
• Oracle RAC database software — Extract the database software into the DB home · Update OPatch in the DB home · Write the DB response file · Install the database software (runInstaller, with the RU) · Run the DB home root.sh (…)
• Creating the RAC database — Create the RAC database (dbca) · Allocate HugePages (…) · Apply the memory settings · Restart the database and check HugePages usage · Verify the database configuration · Check the instances from SQL …
• Service definition and failover tests — Create the application service · Add the client connection definition · Test the connection through SCAN · Failover test: instance crash · Relocate the service back to the preferred instance · Failover test: node loss (stop CRS on the last node) …
• Post-installation — Run the post-installation cluvfy check · Take a manual OCR backup · Configure the basic RMAN settings · Install monitoring and diagnostic tools · Clean up the installation files