wait for reconnect with backoff about franz-go HOT 7 CLOSED

👍 sounds good! Plrease reopen the issue if this isn't worked, I think this is closed.

from franz-go.

There is a function, RetryBackoffFn. The default strategy is to start at 100ms and exponentially backoff with jitter to 1s (this mirrors the behavior described in KIP-144).

I think the timestamp resolution in your logs may not be fine grained enough -- try adding milliseconds with .999 if this is from time.Format.

I'd be opening to changing the default backoff to 250ms rather than 100, and bumping the max backoff to 2.5s or 5s, though. Let me know if upping the time format resolution shows some backoff?

from franz-go.

wow, nice. if i want to not timeout retries and retry always what params i need to specify ?
my use-case if service start consuming and connection broken - retry forever

from franz-go.

Is this for producing, or general admin requests?

Producing already has unlimited retries by default, with acks == -1.
Admin requests have a 1min timeout. I recommend keeping that, but there does exist a RetryTimeout option.

from franz-go.

this question for consuming. so i'm run client.PollFetches and want it connection is dies - auto reconnect to kafka brokers and continue fetching.

from franz-go.

That's the default behavior already, and I think that's the behavior for any client. Any kafka client should continue to reconnect for fetching (with some backoff)

from franz-go.

ok fine, now experimenting with backoff, thanks

from franz-go.