Re: Question about the deprecated policy after 1.0

2016-09-22 Thread Yan Xu
Thanks @haosdent. Sorry there's been offline chats and I was waiting for that to circle back to the list. Comments inlined. Besides this particular issue I think we all agree that we can improve the versioning doc to avoid confusion in the future. I'll send a RR for that but let's focus on

Re: Question about the deprecated policy after 1.0

2016-09-14 Thread haosdent
Thx @yan's update! >The remaining issue we'd like to discuss with the community is, of the things in we plan to support in v1 API and hope to deprecate in Mesos 2.0, should we slate them for deprecation before we have defined their replacement? IMO, if we sure a feature would be deprecated in

Re: Question about the deprecated policy after 1.0

2016-09-14 Thread Yan Xu
To follow up on this, after discussions we (contributors and reviewers of MESOS-6110) have agreed to support the existing HTTP health check API in Mesos 1.x. (See https://reviews.apache.org/r/51803) The remaining issue we'd like to discuss with the community is, of the things in we plan to

Re: Question about the deprecated policy after 1.0

2016-09-06 Thread Yan Xu
Should we think of this *not* as "whether this change should be subject to the current policy" but rather "whether this change presents an *exception case* that *shouldn't* be subject to the current policy"? We shouldn't make exceptions liberally but given that the community as a whole is new to

Re: Question about the deprecated policy after 1.0

2016-09-06 Thread Silas Snider
Responses inline > On Sep 6, 2016, at 1:33 AM, haosdent wrote: > > Hi, Silas. Thanks a lot to help test the health check changes recently. > > According to my understanding about your email, you mentioned two problems: > > 1. The bug that broken exists HTTP/command health

Re: Question about the deprecated policy after 1.0

2016-09-06 Thread haosdent
Hi, Silas. Thanks a lot to help test the health check changes recently. According to my understanding about your email, you mentioned two problems: 1. The bug that broken exists HTTP/command health check caused by r50812 and r50996

Re: Question about the deprecated policy after 1.0

2016-09-05 Thread Silas Snider
There’s a little history to this: In https://reviews.apache.org/r/50812/ , on the 8th of August, the HTTP health check message was changed to be entirely incompatible with the previous HTTP health check message. Not only was its name changed (breaking

Question about the deprecated policy after 1.0

2016-09-05 Thread haosdent
Hi, folks. As I mentioned in the previous email http://search-hadoop.com/m/0Vlr6Ma9DWqzG3M1. We have added `type` in the `HealthCheck` protobuf definition in 1.1.0 and health checks without `type` specified will be deprecated since 1.1.0. For backwards compatibility, we still support the command