Oracle RAC - Zone example.net

category: solutionz · date: 2016-12-31 · updated: 2026-10-02 · author: LALA

Oracle RAC Solution · Config document · referenced from DNS server: Knot, Network and DNS plan

noteThe DBID 1234567890 in the record rac-clsopdb-1234567890 is a made-up value of the right shape, not the identifier of a real database.

The forward zone of the whole installation. It holds one address record for every name the cluster, the Grid Infrastructure installer and Backup Exec use: the public names, the VIPs, the three addresses of the SCAN, the interconnect, backup and management names, the backup server and the DNS server itself.

Item	Value
Path on the host	`/etc/knot/example.net.zone`
Host	`dns1` (`10.30.40.13`), Debian 8.5, Knot DNS
Zone	`example.net`
Serial as built	`2016102609`
Applied with	`/etc/init.d/knot restart`
Checked with	`dig @10.30.40.13 example.net axfr`, 22 records

The file

ini

$ORIGIN example.net.
$TTL 3600
@       SOA     dns1.example.net. hostmaster.example.net. (
                2016102609      ; serial
                6h              ; refresh
                1h              ; retry
                1w              ; expire
                1d )            ; minimum
        NS      dns1

dns1    A       10.30.40.13

; Oracle RAC - public IP
oradb01          A    10.30.10.11
oradb02          A    10.30.10.12

; Oracle RAC - virtual IP
oradb01-vip      A    10.30.10.21
oradb02-vip      A    10.30.10.22

; Oracle RAC - scan IP
oradb-scan       A    10.30.10.31
oradb-scan       A    10.30.10.32
oradb-scan       A    10.30.10.33

; Oracle RAC - interconnect IP
oradb01-int      A    10.30.20.11
oradb02-int      A    10.30.20.12

; Oracle RAC - symantec backend
oradb01-bck      A    10.30.30.11
oradb02-bck      A    10.30.30.12

; Oracle RAC - OS/DB management
oradb01-mng      A    10.30.40.11
oradb02-mng      A    10.30.40.12
oradb01-iscsi    A    10.30.50.11

; Oracle RAC - records for Symantec Backup Exec (the Oracle RAC database named "CLSDB", DBID "1234567890")
rac-clsopdb-1234567890    A    10.30.30.11
rac-clsopdb-1234567890    A    10.30.30.12

; mng-backupsrv01
mng-backupsrv01        A    10.30.40.14
mng-backupsrv01-bck    A    10.30.30.14

Reading the records

Records	Used by
`oradb01`, `oradb02`	Public host names of the nodes, given to the Grid Infrastructure installer as "Public Hostname"
`oradb01-vip`, `oradb02-vip`	Node VIPs, given to the installer as "Virtual Hostname"
`oradb-scan`, three records	The SCAN name; one name that resolves to three addresses
`oradb01-int`, `oradb02-int`	Private interconnect, the alias `eth3:INT` on each node
`oradb01-bck`, `oradb02-bck`	Backup network; the names the Backup Exec agent was installed under
`oradb01-mng`, `oradb02-mng`	Management network
`rac-clsopdb-1234567890`, two records	The virtual RAC node of Backup Exec: database name and DBID, pointing at the backup addresses of both nodes
`mng-backupsrv01`, `mng-backupsrv01-bck`	The Backup Exec server in the management and in the backup network

Three things in the file are not as tidy as they look.

oradb01-iscsi stands under the comment "OS/DB management" although 10.30.50.11 is the iSCSI network, and there is no oradb02-iscsi record at all. The notes give no iSCSI address for the second node anywhere.
The comment above the Backup Exec records names the database CLSDB, while the record itself is built from the name clsopdb. Both names appear in my notes for the same database; see Overview and design.
The NS record and the SOA primary are correct here: dns1 is relative to $ORIGIN example.net. and dns1.example.net. ends with a dot. The reverse zone files do not get this right.

Why the virtual node needs two address records, and what Backup Exec does with them, is in Backup Exec server and RAC backup.

Checked against Knot DNS 3.6.0

As built	Today
Three address records for `oradb-scan`, two for the Backup Exec name, taken to be round robin	Knot 1.6 has no record rotation, so the answers always came in the same order. `answer-rotation` exists since 2.7.3 and is off by default. Oracle's guide still asks for the SCAN addresses to be "configured as round robin addresses" in DNS
One name server, `NS dns1`	RFC 1912 asks for at least two name servers per domain. For a closed internal zone one server answers, but it is a single point of failure
`$TTL 3600` and SOA minimum `1d`	The negative-caching time is the smaller of the SOA record's TTL and its last field, here 3600 seconds and not a day
Serial `2016102609`	Still the recommended form, year, month, day and a two-digit counter
Zone file edited by hand, unsigned	Current Knot can sign zones automatically and can generate reverse records from a forward zone; neither existed in 1.6

What an Oracle client loses when the SCAN name is answered without rotation I could not verify. Oracle's text says "round robin" and does not say what breaks without it.