I've started building a Grafana dashboard to display the stats exposed.
The failure count isn't currently very useful - we can see that a run has failed, but not which job failed etc.
I think it'd be worth re-visiting
assigned to @btasker
The underlying issue is that, when we moved to batched writes (#15) we lost the ability to pass arbitrary arguments in.
We can't simply access class attributes because they might have changed by the time that a batch fails.
What we probably could do though, would be to create a new callback class:
class Callback(object): def __init__(self, config, d): self.job_name = "foo" self.downsample = d def error_callback(self, conf: (str, str, str), data: str, exception) foo
It's attributes could be updated when instantiated to keep track.
The problem with that, though, is that we'd then need a writer handle per job rather than per destination - callbacks are set when opening the API handle.
Activity
10-Feb-23 09:01
assigned to @btasker
10-Feb-23 09:06
The underlying issue is that, when we moved to batched writes (#15) we lost the ability to pass arbitrary arguments in.
We can't simply access class attributes because they might have changed by the time that a batch fails.
What we probably could do though, would be to create a new callback class:
It's attributes could be updated when instantiated to keep track.
The problem with that, though, is that we'd then need a writer handle per job rather than per destination - callbacks are set when opening the API handle.