Achieving data consistency for containers
IBM Storage Fusion backs up application data consistently across all the Persistent Volumes (PVs). It ensures that the recovered data is error-free, reliable, and usable.
The IBM Storage Fusion supports application consistency and crash consistency approaches:
- Application consistency
- In this approach, application I/O must be paused before a snapshot is taken by using recipes. Though it achieves the highest level of consistency, it comes at the cost of application downtime that can be disruptive. Depending on the application, the pause can be for a long time. To know more about how to create and use recipes, see Custom backup and restore workflows.
- Crash consistency
- If IBM Storage Scale is used for storage (and
not Red Hat® OpenShift® Data Foundation), then you can consider crash
consistency. In this approach, application I/O is not paused at all and hence results in zero
disruption. It is achieved by what is called a consistency group. All the PVs of an application are
placed in a consistency group. During backup, a snapshot is taken for all the PVs in the group at
the same instant in time to ensure that all the PVs are consistent with each other.
But
the disadvantage of this approach is that the application might have some data in memory that is not
flushed to the PVs and not included in the snapshot.
In IBM Storage Fusion with IBM Storage Scale, all PVs in a namespace are placed in a consistency group automatically and you do not have to do anything.
Choose the type of data consistency that is appropriate for your application. If the application can restore itself without the data in memory with just what is there in the PVs, then backups are nondisruptive. Otherwise, you must pause I/O and disrupt the application.