feat: more automatic retries #43
Conversation
tustvold
force-pushed
the
retry-handling
branch
from
e52d0aa to
5081e83
Compare
| }) | ||
| .await | ||
| .map_err(Error::Metadata)?; | ||
| let mut backoff = Backoff::new(&self.backoff_config); |
There was a problem hiding this comment.
I initially played around with extracting this retry logic so that it could be shared between here and the PartitionClient, but the different error types made this complicated and so I opted for the simple solution
There was a problem hiding this comment.
We could add a trait for the errors:
trait ErrorHandling {
fn should_invalidate_broker() -> bool;
}
impl ErrorHandling for RequestError {...}
...If I understand the logic correctly this is all you currently need.
There was a problem hiding this comment.
You also need to know what types of errors can be handled , which varies because they are different error enumerations...
I also experimented with just using a single error enumeration for all the clients, this actually works quite nicely but was a substantial amount of churn and I got fed up 😆
|
Created #49 to track improving the test coverage, I agree it should be tested |
tustvold
added
automerge
Configures automatic handling of IO and Poisoned errors