|
|
|
|
|
To use the Secondary in a disaster recovery scenario, the order of writes (write-order fidelity) must be maintained. When replicating in shared disk-group environment, VVR maintains the order of writes by designating one node in the cluster as the logowner. The logowner manages the writes to the SRL on the Primary. The writes are handled differently depending on whether the replication is set to synchronous or asynchronous.
For synchronous RLINKs, all writes are performed on the logowner; writes issued on nodes other than the logowner are sent over the cluster network to the logowner, to be performed there. This process is called write shipping.
For asynchronous RLINKs, the writes are performed on the node where they are issued. However, before writing to the SRL, the node sends a request to the logowner. The logowner responds with a message indicating the position in the SRL that was assigned for that write. After receiving the response from the logowner, the node writes to the SRL and then to the data volumes. This process is called metadata shipping. The information about the position in the SRL and how much space is allocated is known as metadata. If an RVG has both synchronous and asynchronous RLINKs, the RVG uses write shipping.
The logowner also is responsible for replicating the writes for the entire cluster to the Secondary site. If the RLINK is using metadata shipping, the logowner must read back the writes from the SRL before sending the writes to the Secondary.
By default, VVR designates the CVM master as the logowner. It is not required that the master be the logowner. The decision to change the logowner must be made at the cluster-monitor level. The VCS agents for VVR software enables you to change the logowner, if you are using the VCS cluster monitor. You can configure any node in the cluster as the logowner, depending on your requirements. The node that generates the most writes should be configured as the logowner, because the logowner does not have to send messages.
This section explains how VVR processes an incoming write for a Primary cluster containing two nodes. In a shared disk group environment, VVR processes an incoming write on the logowner in the same way as in a private disk group environment.
This example shows how VVR processes an incoming write on the non-logowner for an RVG that is using write shipping. In this example, Node A is the logowner and Node B is the non-logowner.
Example—how VVR processes a write on the non-logowner using write shipping
Click the thumbnail above to view full-sized image.
As shown in the illustration Example—how VVR processes a write on the non-logowner using write shipping, VVR processes an incoming write on the non-logowner (Node B) as follows:
This example shows how VVR processes an incoming write on the non-logowner for an RVG that is using metadata shipping. In this example, Node A is the logowner and Node B is the non-logowner.
Example—How VVR Processes a Write on the Non-logowner Using Metadata Shipping
Click the thumbnail above to view full-sized image.
As shown in the illustration Example—how VVR processes a write on the non-logowner using write shipping, VVR processes an incoming write on the non-logowner (Node B) as follows:
This section explains how VVR processes an incoming read for a Primary cluster containing two nodes. In a shared-disk group environment, VVR processes an incoming read on the master in the same way as in a private disk group environment.
This example shows how VVR processes an incoming read on the non-logowner. In this example, Node A is the logowner and Node B is the non-logowner.
Note
All requests for read and write permissions are sent to the logowner. If the logowner is not the master, it sends permission requests to the master.
Example—how VVR processes a read on the non-logowner
Click the thumbnail above to view full-sized image.
As shown in the illustration, Figure 2-7Example—how VVR processes a read on the non-logowner, VVR processes an incoming read on the non-logowner, (Node B) as follows: