Re: The JIRA Awakens [KAFKA-1841]

2015-12-24 Thread Dana Powers
Awesome ! Looks like a great distribution. I'd love to add HDP releases to the kafka-python integration test suite. -Dana On Dec 24, 2015 9:33 AM, "Harsha" wrote: > HI Dana, > I worked on that release. Yes HDP-2.3.0 has lot of > additional patches on top of 0.8.2.1 mainly

Re: The JIRA Awakens [KAFKA-1841]

2015-12-24 Thread Harsha
HI Dana, I worked on that release. Yes HDP-2.3.0 has lot of additional patches on top of 0.8.2.1 mainly the kerberos patches. We did missed KAFKA-1841 which was fixed in later maint release. We already notified everyone to upgrade HDP-2.3.4 this is the apache ka

Re: Offset Out of Range [Kafka Version:0.8.2.0]

2015-12-24 Thread Fredo Lee
This problem is caused by that the consumer offset lag producer offset, and the messages are deleted, then i use the lagged offset of fetch messages.But those messages have been deleted. Can this be resoled with using `auto.offset.reset=largest` ? 2015-12-24 16:32 GMT+08:00 Fredo Lee : > When u

Re: Protocol version upgrades in 0.9

2015-12-24 Thread Oleksiy Krivoshey
Hi Ewen, Thanks for detailed explanation. So its basically the version of MemberAssignment structure, not the version of the assignment strategy as I thought. Should I use version=0 in protocol exchanges for now? (I'm building a client in Node.js for 0.9: https://github.com/oleksiyk/kafka/blob/mas

Offset Out of Range [Kafka Version:0.8.2.0]

2015-12-24 Thread Fredo Lee
When using kafka-0.8.2.0, Some questions happaned to me. I created one topic called `test` with partitions 60, replicaiton-factors 2 and set log.retention.hours to 24 , Then i send some messages to `test`. some days later, i create a consumer for this topic. but i got `out of range` (i store my of

Re: Protocol version upgrades in 0.9

2015-12-24 Thread Ewen Cheslack-Postava
Oleksiy, The join group protocol is general enough to handle multiple types of group membership, not just consumers. This is used in Kafka Connect to form a group of workers (which, instead of splitting topic partitions between members splits connector tasks). In order to make this work and allow