candumpr is inconsistent with
- does it propagate or suppress errors?
- does it retry or abort?
between the different Sink methods. This is inconsistent, and I don't like it. I'd like to imagine a better strategy.
Perhaps we deactivate the Sink on any error, and retry activating it until the errors go away?
Right now, we drop data in the event of write failures.
candumpr is inconsistent with
between the different
Sinkmethods. This is inconsistent, and I don't like it. I'd like to imagine a better strategy.Perhaps we deactivate the Sink on any error, and retry activating it until the errors go away?
Right now, we drop data in the event of write failures.