Oracle RAC 11g on SUSE Linux - a two-node cluster on Hyper-V

In 2016 I built a two-node Oracle Real Application Clusters database on virtual machines that ran on somebody else's Hyper-V. The database itself was the smaller part of the work. Around it went a DNS server, because the cluster wants its names resolved a certain way, shared disks that had to carry the same name and owner on both nodes, an operating system prepared to the letter of Oracle's installation guide, and a backup product that had to be persuaded that two Linux servers and one database are three things.

This Solution is the whole of that build, written up from my working notes: two SUSE Linux Enterprise Server 11 SP3 nodes with five network roles each, five iSCSI LUNs turned into ASM disks with udev rules and ASMLib, Oracle Grid Infrastructure and Oracle Database 11.2.0.4 Standard Edition, an authoritative Knot DNS server on Debian, and Symantec Backup Exec 15 backing up the nodes and the database. Two things went wrong on the way and both are told in full: the root script of the Grid installation failed on this SUSE service pack until a patch was applied, and the second node could not join the cluster because the virtual network passed neither multicast nor broadcast.

The system in one picture

flowchart TB
  apps["Database clients"]
  subgraph cluster["Oracle RAC, cluster clsdb"]
    scan["SCAN oradb-scan, three addresses"]
    n1["oradb01: Grid Infrastructure, ASM, instance 1"]
    n2["oradb02: Grid Infrastructure, ASM, instance 2"]
    n1 <-- "interconnect, alias eth3:INT" --> n2
  end
  subgraph san["Shared iSCSI LUNs"]
    crs["ORADB-CRS: disk group OCR"]
    data["ORADB-DATA01 and DATA02: disk group DATA"]
    fra["ORADB-FRA: disk group FRA"]
    ontrlg["ORADB-ONTRLG: disk group ONTRLG"]
  end
  dns["dns1: Knot DNS and NTP"]
  be["mng-backupsrv01: Backup Exec server"]
  apps -- "public network" --> scan
  scan --> n1
  scan --> n2
  n1 -- "iSCSI network" --> san
  n2 -- "iSCSI network" --> san
  n1 -- "management network" --> dns
  n2 -- "management network" --> dns
  be -- "backup network" --> n1
  be -- "backup network" --> n2

The fictional environment

Every Article and every Config document uses the same names and addresses.

Host, cluster, database, disk and account names are role names and were kept. The SCSI identifiers, the database identifier and the voting file identifier in the listings are made up.

Articles

Read in this order; it is the order the system was built in.

Configuration

Each document holds one file, one command set or one set of installer answers as it was used, with comments, the differences between the two nodes, and a check against the current release. Oracle's installers and the Backup Exec console leave no file to show, so their documents list what was answered, screen by screen.

DNS

Operating system

Storage

Grid Infrastructure

Database

Backup

What the write-up found

Reading the notes again ten years later turned up things I did not see when I wrote them. Each is told in its Article.