What can we do to prevent kafka outages?

[He-Pin]

https://www.pagerduty.com/eng/august-28-kafka-outages-what-happened-and-how-were-improving/

[He-Pin]

Usage of the pekko-connectors-kafka Scala library – passing Kafka settings instead of a pre-created producer instance led the connector to generate a new producer per request rather than reusing an existing one as intended.

refs: https://pekko.apache.org/docs/pekko-connectors-kafka/current/send-producer.html

[He-Pin]

https://pekko.apache.org/docs/pekko-connectors-kafka/current/send-producer.html

[pjfanning]

Why do we need #361 and #362?
Can we get a reproducible test case? This is a volunteer project and it is much simpler if people reporting issues do some of the leg work.

[He-Pin]

I think they encounter this problem because they create too many Producer, and the Producer is a case class, which cause they more time to spot , because of the missing new keword.

[pjfanning]

Can we just close this and let PagerDuty report this if they care? The blog seems to say they fixed the issue themselves.
If they didn't raise an issue with us then I'm simply going to assume that they misconfigured something.

[He-Pin]

I'm ok. I have some idea about this.