Can it be called a pure AP system if a long running network partition can cause it to become unavailable?
In our case, shard groups are part of the data that is in the CP system. We create these ahead of time, but those are what we use to determine where in the cluster a given write should go and what servers we need to hit to answer a query.
Let's call a "normal partition" one that is less than a few hours. In that case I wouldn't expect it to cause the system to become unavailable. However, there are certainly scenarios in which a longer partition would make it unavailable.
And even for a "normal partition", new databases wouldn't be available to write to.
In our case, shard groups are part of the data that is in the CP system. We create these ahead of time, but those are what we use to determine where in the cluster a given write should go and what servers we need to hit to answer a query.
Let's call a "normal partition" one that is less than a few hours. In that case I wouldn't expect it to cause the system to become unavailable. However, there are certainly scenarios in which a longer partition would make it unavailable.
And even for a "normal partition", new databases wouldn't be available to write to.