Details
-
Improvement
-
Status: Resolved
-
Minor
-
Resolution: Fixed
-
1.1.1
Description
Offsets for a given partition at a particular timestamp can now be found using offsetsForTimes API. https://kafka.apache.org/0110/javadoc/org/apache/kafka/clients/consumer/KafkaConsumer.html#offsetsForTimes(java.util.Map).
One way to make use of this api would be to :
Add a new option for FirstPollOffsetStrategy called TIMESTAMP
Add a new startTimeStamp option to KafkaSpoutConfig, which would be used only when FirstPollOffsetStrategy is set to TIMESTAMP.
Later in the KafkaTridentSpoutEmitter, when we do the first seek, we can do something like :
if(firstPollOffsetStrategy.equals(TIMESTAMP)) { try { startTimeStampOffset = kafkaConsumer.offsetsForTimes(Collections.singletonMap(tp, startTimeStamp)).get(tp).offset(); } catch (IllegalArgumentException e) { LOG.error("Illegal timestamp {} provided for tp {} ",startTimeStamp,tp.toString()); } catch (UnsupportedVersionException e) { LOG.error("Kafka Server do not support offsetsForTimes(), probably < 0.10.1",e); } if(startTimeStampOffset!=null) { LOG.info("Kafka consumer offset reset for TopicPartition {}, TimeStamp {}, Offset {}",tp,startTimeStamp,startTimeStampOffset); kafkaConsumer.seek(tp, startTimeStampOffset); } else { LOG.info("Kafka consumer offset reset by timestamp failed for TopicPartition {}, TimeStamp {}, Offset {}. Restart with a different Strategy ",tp,startTimeStamp,startTimeStampOffset); } }
Attachments
Issue Links
- is duplicated by
-
STORM-2992 Add support for starting storm-kafka-client spout consumer at a specific timestamp
- Closed
- links to