Using Filebeat v1.2.3:
In case of error (no correct ACK received), filebeat retries to send the frame but before, the process is stopped during a computed duration: backoff (that is exponential until 1min) . But after an ACK is sent (no more problem) the value is not reset. So in case of new error, backoff will be probably 1min (until a filebeat reboot).
Maybe we can reset backoffCount that is used to compute backoff? check: https://github.com/adioss/beats/commit/793065013faedfaff030641988657b33c4076df4
Thaks for reporting. Definitely a bug in 1.2.x release. Can you create an github issue for tracking this? Proposed change is incomplete. The variable backoffCount should be reset at line 135 too. Backoff behavior has been fixed/changed in master long time ago.
Thanks for ticket plus pull request. I can set labels. But you can not set v1.2.3 as roadmap is still up to us (and 1.2.3 already being released). We might prepare for a new 1.x version soonish, so we have to wait for PR potentially going into another branch.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.