Resolving object differences between sites

Objects can be synchronized in two ways.

The process of object synchronization is used in both standard vault mirrors and protected mirrors, and the process is the same regardless of the mirror type.

When objects are written, modified, or deleted, the operation is attempted on both vaults of a mirror in parallel. If the write to both vaults are not successful, synchronization operations are used to synchronize the two vaults. Objects can be synchronized in two ways:

Slicestor® Node Synchronization

Background Slicestor® Node-based scanning and synchronizing processes are used. A Slicestor® Node in a vault mirror constantly runs a scanning process that scans random ranges within their mirrors in parallel on both vaults. Any object written in the past hour will be skipped to avoid I/O operations in progress.

Accesser® Node Synchronization

If the object was written recently, the objects are synchronized using an Accesser® Node device-based synchronization. A queue on the Accesser® Node records all objects written to one vault that could not be written to the other vault. The Accesser® Node only has capacity for a limited number of objects to be synchronized. This number is very small. This mechanism is not intended to handle extended network partitions.

The Accesser® Node-level synchronization process will write the second copy if either:

• The mirror state is Good, or
• An hour has passed.