Database Reference
In-Depth Information
an available interface. Output of ifconfig shows that logical interface eth3:1 is configured with link local -IP address
169.254.28.111 and that eth4:1 is configured with link local IP address 169.254.78.12. Both Database and Clusterware
will use both IP addresses for private network communication.
$ ifconfig -a |more
...
eth3: ..<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 9000 index 4
inet 172.18.1.13 netmask ffffff00 broadcast 172.18.1.255
eth3:1: ..<UP,BROADCAST,RUNNING,MULTICAST,IPv4 > mtu 9000 index 4
inet 169.254.28.111 netmask ffffc000 broadcast 169.254.63.255
...
Eth4: ..<UP,BROADCAST,RUNNING,MULTICAST,IPv4> mtu 9000 index 4
inet 172.18.2.15 netmask ffffff00 broadcast 172.18.2.255
eth4:1: ..<UP,BROADCAST,RUNNING,MULTICAST,IPv4 > mtu 9000 index 4
inet 169.254.78.12 netmask ffffc000 broadcast 169.254.127.255
...
You can identify the IP addresses and ports used by a process for cache fusion traffic using the
oradebug ipc
command. In the following example, I used the oradebug command to dump IPC context information to a trace file
after connecting to the database.
RS@RAC1>oradebug setmypid
Oracle pid: 6619, Unix process pid: 28096, image: oracle@node1
RS@RAC1>
oradebug ipc
Information written to trace file.
RS@RAC1>oradebug tracefile_name
/opt/app//product/12.1.0/admin/diag/rdbms/RAC/RAC1/trace/RAC1_ora_28096.trc
From the trace file, you can identify that foreground process listens in four IP addresses. Foreground process uses
all four UDP ports to communicate between processes in other nodes. Essentially, since all IP addresses are in use,
this scheme provides load balancing among the IP addresses.
SKGXP: SKGXPGPID Internet address 168.254.28.111 UDP port number 45913
SKGXP: SKGXPGPID Internet address 168.254.78.12 UDP port number 45914
SKGXP: SKGXPGPID Internet address 168.254.36.6 UDP port number 45915
SKGXP: SKGXPGPID Internet address 168.254.92.33 UDP port number 45916
If there is an interface assigned for cluster interconnect, then the HAIP feature will configure one link local IP
address in that interface. If there are two interfaces assigned for private network, then two link local IP addresses will
be configured per node. If more than two interfaces are assigned, then four link local IP addresses will be configured
in those interfaces per node. For example, if there are four interfaces configured in the database server for private
network, then Clusterware will configure one link local IP address each in the configured interfaces.
Also, Clusterware uses the network configuration from the first node you execute, root.sh, as the reference
configuration. So, it is important to keep network configuration consistent in all nodes of a cluster.
