The application developer can setup priority on transactions. These priorities have huge effect on how these transactions perform and fail, which is the desired effect.
Especially batch transactions can be suspended entirely if the cluster is under load. For example, data movement caused by host failure. In some cases, this can be suspended very long as the data movement lasts 10s of minutes.
We have some documentation on API docs and transaction options. We should make this more explicit.
This is documented on the transaction options, but I think the desire is that it be made clear that the cutting off of transactions can happen for extended periods of time.
The application developer can setup priority on transactions. These priorities have huge effect on how these transactions perform and fail, which is the desired effect.
Especially batch transactions can be suspended entirely if the cluster is under load. For example, data movement caused by host failure. In some cases, this can be suspended very long as the data movement lasts 10s of minutes.
We have some documentation on API docs and transaction options. We should make this more explicit.