Closed horkhe closed 9 years ago
After a Kafka node crash consume requests started to return 408 Request Timeout even though there were messages in the queue. During the incident the following line appeared over and over in the logs:
</offsetManager[2]/kraken:ml_count_events2:1[26]/processCommits[0]> failed to fetch initial offset: err=(write tcp 10.223.192.38:48356->10.223.144.128:9092: write: broken pipe)
This seemed to be a bug in the OffsetManager implementation of the Mailgun clone of Shopify/sarama.
After a Kafka node crash consume requests started to return 408 Request Timeout even though there were messages in the queue. During the incident the following line appeared over and over in the logs: