I think some of that deals with the fact that the error only occurs on worker startup, but it presents the possibility, although extremely rare, that a tuple tree may be marked as fully processed by an acker when it has not been. It also violates one of storms guarantees. I would rather wait, but I am willing to let 0.9.2 out with a -0 because I can see an argument for STORM-342 just being critical. For me personally I plan to deploy a new release with STORM–342 to the clusters I manage ASAP.
- Bobby From: "P. Taylor Goetz" <[email protected]<mailto:[email protected]>> Reply-To: <[email protected]<mailto:[email protected]>> Date: Thursday, June 12, 2014 at 1:56 PM To: <[email protected]<mailto:[email protected]>> Subject: Re: 0.9.2-incubating Release Pending I agree. The fact that this was submitted as a blocker slipped past me. In my opinion blockers are serious “stop the bus — this software is fundamentally broken” issues. This bug has presumably been present in every release since the move to disruptor way back in 0.8.0, and only discovered/reported now (excellent catch Sean). So it doesn’t feel like a blocker. That being said, I think it’s a very important issue to get resolved. I’m willing to cancel the current 0.9.2 release vote (there aren’t any votes yet), review the patch, and re-release if that’s how we want to proceed. What do you think? - Taylor On Jun 12, 2014, at 2:00 PM, Bobby Evans <[email protected]<mailto:[email protected]>> wrote: I don¹t like the idea of releasing code with a known blocker in it. Either 0.9.2 needs STORM-342 or it is not a blocker. I don¹t see how we can have it both ways. - Bobby On 6/11/14, 2:47 AM, "Sean Zhong" <[email protected]<mailto:[email protected]>> wrote: It make sense, thanks! Sean On Wed, Jun 11, 2014 at 9:25 AM, P. Taylor Goetz <[email protected]<mailto:[email protected]>> wrote: Agreed. But the patch needs to be reviewed and signed off. If it is genuinely reproducible, and fixed with the patch, then I'm not opposed to doing a rapid follow-on release of 0.9.3-incubating. But for now, there are a lot of important patches in 0.9.2-incubating (yours included) that the community has been waiting a long time for. I would like to keep the pace of releasing at least quarterly, if not more often. -Taylor On Jun 10, 2014, at 8:50 PM, Sean Zhong <[email protected]<mailto:[email protected]>> wrote: How about STORM-342? STORM-342: Message loss, executor hang, or message disorder due to contention in Disruptor queue under multi-thread mode https://issues.apache.org/jira/browse/STORM-342? It looks like a big issue to me. On Wed, Jun 11, 2014 at 3:58 AM, P. Taylor Goetz <[email protected]<mailto:[email protected]>> wrote: Unfortunately RAT turned up a bunch of licensing issues, so there will be a delay. I expect to have a release cut and ready for vote sometime tomorrow. - Taylor On Jun 9, 2014, at 6:03 PM, P. Taylor Goetz <[email protected]<mailto:[email protected]>> wrote: FYI... I will cut a release candidate tomorrow. -Taylor On May 29, 2014, at 3:01 PM, "P. Taylor Goetz" <[email protected]<mailto:[email protected]>> wrote: I¹m planning on cutting a release candidate for 0.9.2 in the next day or so. I¹d like to ask committers to review outstanding contributions that I¹d like to see in the release. For now I only see one remaining pull request that I¹d like to include: STORM-205 (Rest API for Storm UI) I¹m also open to suggestions for additional JIRAs to include. Thanks in advance, - Taylor
