Hi, I'm not sure that this is a good solution. What happens when failOnError is set to FALSE, but the deliveryGuarantee is set to EXACTLY_ONCE?
Best regards, Martijn On Sun, Aug 24, 2025 at 5:57 PM Ejaskhan S <[email protected]> wrote: > Hi All, > > Any feedback on this ? > > Best regards. > Ejaskhan > > > > > On Wed, Aug 20, 2025 at 4:46 PM Ejaskhan S <[email protected]> wrote: > > > Hi All, > > > > Any feedback on this ? > > > > Best regards. > > Ejaskhan > > > > > > > > > > On Tue, Aug 12, 2025 at 9:50 AM Ejaskhan S <[email protected]> > wrote: > > > >> Hi Flink Devs, > >> > >> *Description:* > >> Currently, even when DeliveryGuarantee is set to NONE, the Kafka sink > >> still throws exceptions (for example, due to Kafka cluster issues), > which > >> in turn causes the Flink job to restart. This behavior seems > >> counterintuitive for NONEguarantee mode, where the expectation is > >> typically to avoid failures that impact job continuity. > >> > >> *Proposal:* > >> I would like to propose introducing a new configuration parameter, > >> failOnError, in KafkaSinkBuilder. > >> > >> - > >> > >> When failOnError is set to true (default), the existing behavior > >> remains unchanged—exceptions from the Kafka cluster will cause the > job to > >> fail and restart. > >> - > >> > >> When failOnError is set to false, the sink would log the error and > >> continue without throwing an exception, thus avoiding job restarts. > >> > >> This change would give users more control over failure handling, > >> especially in scenarios where the delivery guarantee is already NONE and > >> uninterrupted processing is preferred over strict failure enforcement. > >> > >> I have already created a Jira issue for this: FLINK-38226 > >> <https://issues.apache.org/jira/browse/FLINK-38226> and I am ready to > >> work on the implementation if it is assigned to me. > >> > >> Looking forward to your thoughts and feedback on whether this change > >> would be useful for the broader community. > >> > >> Best regards, > >> Khan > >> > >> > >> > >> > >> On Tue, Aug 12, 2025 at 9:38 AM Ejaskhan S <[email protected]> > wrote: > >> > >>> Hi Flink Devs, > >>> > >>> *Description:* > >>> Currently, even when DeliveryGuarantee is set to NONE, the Kafka sink > >>> still throws exceptions (for example, due to Kafka cluster issues), > which > >>> in turn causes the Flink job to restart. This behavior seems > >>> counterintuitive for NONEguarantee mode, where the expectation is > >>> typically to avoid failures that impact job continuity. > >>> > >>> *Proposal:* > >>> I would like to propose introducing a new configuration parameter, > >>> failOnError, in KafkaSinkBuilder. > >>> > >>> - > >>> > >>> When failOnError is set to true (default), the existing behavior > >>> remains unchanged—exceptions from the Kafka cluster will cause the > job to > >>> fail and restart. > >>> - > >>> > >>> When failOnError is set to false, the sink would log the error and > >>> continue without throwing an exception, thus avoiding job restarts. > >>> > >>> This change would give users more control over failure handling, > >>> especially in scenarios where the delivery guarantee is already NONE > and > >>> uninterrupted processing is preferred over strict failure enforcement. > >>> > >>> I have already created a Jira issue for this: KAFKA-19547 > >>> <https://issues.apache.org/jira/browse/KAFKA-19547?filter=-2>, and I > am > >>> ready to work on the implementation if it is assigned to me. > >>> > >>> Looking forward to your thoughts and feedback on whether this change > >>> would be useful for the broader community. > >>> > >>> Best regards, > >>> Khan > >>> > >> >
