How geographic mirroring works
This topic describes how geographic mirroring works.
Configuring
The nodes participating in geographic mirroring must be in the same cluster, the same device domain, and the same cluster resource group. Before configuring geographic mirroring, specify a site name and the TCP/IP address(es) for each node in the recovery domain. If you have more than one node at a site, then the hardware (disk units) you select for the disk pool must be switchable between the nodes at the site. If you only have one node at a site, the hardware does not have to be switchable and should be non-switchable (private).
When geographic mirroring is configured, the mirror copy has the same disk pool number and name as the original disk pool, the production copy. Geographic mirroring is logical mirroring, not physical mirroring. The two disk pools must have similar disk capacities, but the mirror copy may have different numbers and types of disk units as well as different types of disk protection.
Managing
After geographic mirroring is configured, the production copy and mirror copy function as a disk unit. When the production copy is made available, the mirror copy is brought to a state that allows geographic mirroring to be performed. Synchronization occurs when you make the disk pool available after you configure geographic mirroring. When geographic mirroring is active, changes to the production copy data are transmitted to the mirror copy across TCP/IP connections. Changes can be transmitted either synchronously or asynchronously.
- Synchronous mode
- The client waits until the operation is complete to disk on both the source and target systems. The mirror copy is always eligible to become the production copy, because the order of writes is preserved on the mirror copy. IBM recommends to try synchronous mode first. If your performance remains acceptable, continue to use synchronous mode.
- Asynchronous mode
- The client must wait only until the operation is complete to disk on the source system and is received for processing on the target system. However, synchronous mode is safer because if the primary node fails or the production copy fails, the mirror copy can become the production copy. In asynchronous mode, the pending updates must be completed before the mirror copy can become the production copy.
To maintain the data integrity of the mirror copy, the user cannot access the mirror copy while geographic mirroring is being performed. The user can detach the mirror copy to perform save operations, to create reports, and to perform data mining. However, the mirror copy must be synchronized with the production copy after it is reattached.
Tracking space
To suspend geographic mirroring with tracking, you set the tracking space when you configure geographic mirroring or change geographic mirroring attributes. Tracking space is allocated in the independent ASPs. The more tracking space you specify, the more changes the system can track. The maximum tracking space allowed is approximately 1% of the independent ASPs capacity.
Suspending geographic mirroring with tracking
If you suspend with geographic mirroring with tracking, the system will attempt to track changes made to those disk pools. This may reduce the synchronization process by performing partial synchronization when you resume geographic mirroring. If tracking space is exhausted, then when you resume geographic mirroring, complete synchronization is required.
When you resume geographic mirroring, a complete synchronization can be a lengthy process, anywhere from several hours to even longer.
Suspending without tracking
If you suspend geographic mirroring without tracking changes, then when you resume geographic mirroring, a complete synchronization is required between the production and mirror copies. If you suspend geographic mirroring and you do track changes, then only a partial synchronization is required. Complete synchronization can be a very lengthy process, anywhere from one hour, to several hours, to even longer. The length of time it takes to synchronize is dependent on the number and type of disk units as well as how many TCP/IP communication interfaces are dedicated to geographic mirroring.
Synchronization
The production copy can function normally during synchronization, but performance might be negatively affected. During synchronization, the contents of the mirror copy are unusable, and it cannot become the production copy. If the independent disk pool is made unavailable during the synchronization process, synchronization resumes where it left off when the independent disk pool is made available again.
The first % complete message (CPI095D), after resuming an interrupted synchronization, shows 0%.
Synchronization type
These are two types of synchronization:
Full synchronization
- Indicates that a complete synchronization takes place. Changes to the production copy are not tracked to apply to the synchronization.
- Deletes all of the data on the mirror copy and copies all of the latest data from the production copy to the mirror copy.
Partial synchronization
- Indicates that changes to the production copy are tracked to apply to the synchronization. This may shorten the synchronization time because a complete synchronization is unnecessary.
Synchronization priority
When you set the attributes for geographic mirroring, you can set the synchronization priority. If synchronization priority is set high, the system uses more resources for synchronization, which results in a sooner completion time. The mirror copy is eligible to become a production copy faster, so you are protected sooner. However, high priority can cause degradation to your application. IBM recommends that you try high priority first, so you are protected as soon as possible. If the degradation to your application performance is not tolerable, then lower the priority.
Recovery timeout
In addition to synchronization priority, you can also set recovery time out. The recovery timeout specifies how long your application can wait when geographic mirroring cannot be performed. When an error, such as IP failure, prevents geographic mirroring, the source system waits and retries for the specified recovery timeout before suspending geographic mirroring which allows your application to continue. The trade-off is between blocking your application or requiring synchronization after suspending geographic mirroring. When your application is blocked for an extended time, other jobs might also be blocked waiting for resources and locks owned by the applications using the geographic mirrored disk pool. When geographic mirroring is suspended, you no longer have the protection of the mirror copy. If your application can tolerate a delay, it is recommended to set recovery timeout from 2 to 5 minutes. If the volume of your data is large (over a terabyte), consider a longer recovery timeout value to reduce the possibility of suspending geographic mirroring. If mirroring is suspended without tracking, the system performs a full synchronization. If geographic mirroring is suspended with tracking, the system performs a partial synchronization.
System roles
When you configure the cluster for geographic mirroring, you have many options for defining the availability and protection of the independent disk pool. When you create the switchable hardware group, you list the order of the backup systems to which the independent disk pool will failover or switch over. If the primary node switches to a backup node at the same site, a hardware switch will occur. If the primary node switches to the other site, the mirror copy on the backup node changes roles to become the production copy. The old primary node becomes the new backup node, and the production copy becomes the mirror copy. The new production copy is now accessible for updates on the remote system. If the independent disk pools are part of a disk pool group, all of the disk pools in the group will switchover together. See Example: Independent disk pools with geographic mirroring.
Parent topic:
Geographic mirroring concepts