Read conditions for unsynchronized objects
If, during a network partition, an object is updated in both vaults independently, it is unknown which object will be read once the network is restored.
- Requesting an object written to only one site will return the object if both sites are accessible.
- Requesting an object deleted from one site but not the other will return the object if both sites are accessible.
- If a network partition exists between the two sites, a client can read any object on their side of that partition.
- If an object does not exist and both sites are accessible, the request will not complete
until both sites have responded. CAUTION:This results in higher latency for the object not found condition since both vaults in a mirror must report that the object does not exist. This is done to ensure an object exists whenever possible. See Standard and protected vault mirrors favor availability (returning an object) for more information