Physical IP, virtual IP and Scan IP in RAC

Using physical IP:

----------------------

  If the client request for connection goes to Physical IP of a node and the node1 is down, client will get notified about after network time out only (which could be as long as 10 minutes). After the client gets this information, he tries to connect to the alternative address node2.

Con: takes time to throw error so takes time to reconnect.

 

Using virtual IP:

-------------------

Why must we use a VIP address in an Oracle RAC environment? 

The simple answer to this question is to avoid TCP timeouts.

  If the client request goes to Virtual IP of the node (VIP) and the node is down, VIP will automatically failover to one of the surviving nodes without waiting for the network time-out and client gets connected.

server node fail, this VIP is transferred to another server node in order to still provide network connectivity for clients using the address of the failed server node. VIP in other words provides high availability as despite a server node failing, network communication to this node will still be supported by another node via the failed node's VIP.

Con:
On a 10 node cluster, if there will be 10 virtual IP addresses with 10 virtual hostnames - which means that many clients will need to know and use all 10 VIPs in order to make load balanced, or high availability, or TAF, connections.

 

Using SCAN IP:

--------------------
SCAN will do this on the client side - by providing the client with a Single Client Acces Name to use as oppose to 10 VIPs. The SCAN listener service knows all cluster services - and can redirect the client to the appropriate VIP listener where the requested service is available.

Scan has 3 IPs configured in DNS server instead of one IP because to distribute total load on single IP.

Oracle recommends 3 IPs for scan is enough for load distribution among 3 scan IPs and provides ease in management of cluster nodes (addition or deletion).

You can add new nodes to the cluster without changing your client TNSnames. This is because Oracle will automatically distribute requests accordingly based on the SCAN IPs which point to the underlying VIPs. Scan listeners do the bridge between clients and the underlying local listeners which are VIP-dependent.



Suppose, we have a two node Real Application Cluster set up with following IP's

NODE              Static IP address                     Virtual IP address
==============================================
racnode1            192.168.1.100                         192.168.1.200 
                     (racnode1)                           (racnode1_vip1)

racnode2           192.168.1.101                          192.168.1.201 
                    (racnode2)                            (racnode2_vip2)


In Database Management Software Oracle 10g:

Let's first see how this works in Oracle 10g. Suppose, Listener.ora of both Database is using Static IP for it's configuration like 

LISTENER=
  (DESCRIPTION=
    (ADDRESS_LIST=
      (ADDRESS=(PROTOCOL=tcp)(HOST=racnode1)(PORT=1521))
      (ADDRESS=(PROTOCOL=ipc)(KEY=extproc))))
SID_LIST_LISTENER=
  (SID_LIST=
    (SID_DESC=
      (GLOBAL_DBNAME=sales.us.example.com)
      (ORACLE_HOME=/oracle10g)
      (SID_NAME=Service1))
    (SID_DESC=
      (SID_NAME=plsextproc)
      (ORACLE_HOME=/oracle10g)
      (PROGRAM=extproc)))


Hence, Tnsnames.ora for Client system will be like 

Service1 =
(DESCRIPTION =
  (ADDRESS=(PROTOCOL=TCP)(HOST=racnode1)(PORT=1521))
  (ADDRESS=(PROTOCOL=TCP)(HOST=racnode2)(PORT=1521))
    (CONNECT_DATA =
      (SERVICE_NAME = Service1)
     )
  )

Now, A new connection to database will first go to racnode1, if this node is alive and working fineconnection will be establish and user can continue work.

What if, racnode1 is not available Even in this case, client tries to establish a connection with theracnode1 Because, it is first in its address list.But since the node(racnode1) is not available, client tries to establish it’s connection with the next available address in the list (i.e racnode2). So, there is a delay to move from one node to other. This is called Connect-Time Failover. 

But the Problem is that the TIME (TCP TIMEOUT) it takes to failover, which will be ranging between a few seconds to a few minutes. For a very high critical systems/environments this is not acceptable.

To resolve this problem Oracle introduce Virtual IP (VIP).

Let's see how it works with VIP.

Now, Listener.ora of both Database is using VIP for it's configuration like 

LISTENER=
  (DESCRIPTION=
    (ADDRESS_LIST=
      (ADDRESS=(PROTOCOL=tcp)(HOST=racnode1_vip1)(PORT=1521))
      (ADDRESS=(PROTOCOL=ipc)(KEY=extproc))))
SID_LIST_LISTENER=
  (SID_LIST=
    (SID_DESC=
      (GLOBAL_DBNAME=sales.us.example.com)
      (ORACLE_HOME=/oracle10g)
      (SID_NAME=Service1))
    (SID_DESC=
      (SID_NAME=plsextproc)
      (ORACLE_HOME=/oracle10g)
      (PROGRAM=extproc)))


Hence, Tnsnames.ora for Client system will be like 

Service1 =
(DESCRIPTION =
  (ADDRESS=(PROTOCOL=TCP)(HOST=racnode1_vip1)(PORT=1521))
  (ADDRESS=(PROTOCOL=TCP)(HOST=racnode1_vip2)(PORT=1521))
    (CONNECT_DATA =
      (SERVICE_NAME = Service1)
     )
  )

Now, A new connection to database will first go to racnode1_vip1, if this node is alive and workingfine connection will be establish and user can continue work.

What if, racnode1_vip1 is not available Even in this case, client tries to establish a connection with the racnode1_vip1 Because, it is first in its address list. But since the node(racnode1_vip1) is notavailable, CRS will come in to picture and move the failed node’s VIP to one of the surviving nodes of the cluster.

 Any connection attempts to the failed node by using VIP will be handled by the failed node’s VIP that is currently residing on one of the surviving node. 

This (failed node’s VIP) will respond immediately to client by sending an error indicating that there is no listener. Upon receiving the information of no listener,client immediately retry connection using the next IP in the address list. Thus reduces the time to failover. 

In Database Management Softwere Oracle 11g2:

When we talk about Oracle 11g, Since, we have SCAN VIP's in Oracle 11g, Following question comes into mind 
  
Do we still need VIP in Oracle 11g ?

 Yes, We still need VIP. VIP still play the same role as it is discussed in case of Database Management Softwere Oracle 10g.

 What is the Difference between SCAN VIP and VIP ?

The IP address corresponding to SCAN NAME are called as SCAN VIP. Which runs on DB nodes as SCAN LISTENERS.  

 Let's see how VIP's works in 11g R2.  In Oracle 11g R2, tnsnames.ora will have only one entry that is scan name of the Cluster like.

Service1 =
(DESCRIPTION =
  (ADDRESS=(PROTOCOL=TCP)(HOST=scan_racnode1_vip1)(PORT=1521))
    (CONNECT_DATA =
      (SERVICE_NAME = Service1)
     )
  )

This scan name is resolved by any of the SCAN VIP and every SCAN VIP has a Listener associated with it running on node know as SCAN LISTENER. In below example, There are two SCAN LISTENER's running on odain1 and odain2.

[grid@bin]$ srvctl status scan_listener
SCAN Listener LISTENER_SCAN1 is enabled
SCAN listener LISTENER_SCAN1 is running on node odain1
SCAN Listener LISTENER_SCAN2 is enabled
SCAN listener LISTENER_SCAN2 is running on node odain2

All databases are registered with each SCAN LISTENER in the Cluster and PMON updates it's load to each SCAN LISTENER. Each request go through using SCAN_NAME, resolves to SCAN VIP i.e. SCAN LISTENER. Now, SCAN LISTENER redirects it to VIP by deciding using Load Balance.

SCAN _NAME ===============> SCAN VIP ==============> VIP