Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Kafka, mail # user - status of 0.8?


Copy link to this message
-
Re: status of 0.8?
graham sanderson 2013-02-22, 05:37
thanks Jun

On Feb 21, 2013, at 11:26 PM, Jun Rao <[EMAIL PROTECTED]> wrote:

> The remaining 0.8 blockers are tracked in
> https://cwiki.apache.org/confluence/display/KAFKA/Kafka+replication+development
>
> Thanks,
>
> Jun
>
> On Thu, Feb 21, 2013 at 9:14 PM, graham sanderson <[EMAIL PROTECTED]> wrote:
>
>> Thanks, so if I understand what you are saying is that 0.8 is alpha-ish;
>> i.e. everything pretty much implemented (is there anything major missing -
>> hard to tell looking at JIRA without much context) - we use at our own risk
>> (which we're fine with), but that given we really want 0.8 features we can
>> and should start testing with it today, with an expectation of a release in
>> the next N months where N is hopefully no more than 2
>>
>> On Feb 21, 2013, at 11:06 PM, Neha Narkhede <[EMAIL PROTECTED]>
>> wrote:
>>
>>> 0.8 is a significant rewrite since the 0.7 codebase. Hence, there is no
>>> easy way to "turn off" replication and expect 0.7.2 like functionality.
>>> Having said that, it is somewhat stable so you can use it for
>> prototyping.
>>>
>>> Thanks,
>>> Neha
>>>
>>>
>>> On Thu, Feb 21, 2013 at 1:42 PM, graham sanderson <[EMAIL PROTECTED]>
>> wrote:
>>>
>>>> So I am excited that 0.8 is getting close.
>>>>
>>>> I understand that 0.8 is pre-beta, but to what extent would you say the
>>>> bugs are mostly with the new redundancy code vs regressions in existing
>>>> functionality;  i.e. if wanting to prototype and deploy something over
>> the
>>>> next month or so with the view to deploying it within the next two
>> months,
>>>> is it safe-ish to go with 0.8 with replicated brokers turned off (for
>> now)
>>>> vs going with 0.7.2?
>>>>
>>>> Also does anyone know the status of any node.js clients for 0.8
>>>>
>>>> On Feb 21, 2013, at 3:32 PM, David Arthur <[EMAIL PROTECTED]> wrote:
>>>>
>>>>> I've been having trouble building trunk since the changes to sbt in
>> 0.8.
>>>> Is there some documentation on building and running trunk?
>>>>>
>>>>>
>>>>> On 2/21/13 3:53 PM, Neha Narkhede wrote:
>>>>>> HEAD is good as of today and has been stable for past few days. There
>>>> are
>>>>>> some bugs we are working on but you can certainly run the cluster and
>> do
>>>>>> some basic send/receive operations. Also, let us know if you have
>>>> feedback
>>>>>> on the APIs, protocols, tools etc since that takes some time to
>> refactor
>>>>>> and change.
>>>>>>
>>>>>> Good luck! :-)
>>>>>>
>>>>>> Thanks,
>>>>>> Neha
>>>>>>
>>>>>>
>>>>>> On Thu, Feb 21, 2013 at 12:51 PM, Jason Rosenberg <[EMAIL PROTECTED]>
>>>> wrote:
>>>>>>
>>>>>>> Thanks Neha,
>>>>>>>
>>>>>>> That's helpful info.  Is there a reasonable checkpoint rev to check
>>>> out now
>>>>>>> and experiment with, or is HEAD as good as anything else?
>>>>>>>
>>>>>>> Jason
>>>>>>>
>>>>>>> On Thu, Feb 21, 2013 at 3:31 PM, Neha Narkhede <
>>>> [EMAIL PROTECTED]
>>>>>>>> wrote:
>>>>>>>> Hi Jason,
>>>>>>>>
>>>>>>>> We are closely monitoring the health of one of our production
>> clusters
>>>>>>> that
>>>>>>>> has the 0.8 code deployed. This cluster is feeding off of LinkedIn's
>>>>>>>> production traffic. Once this cluster is fairly stable, we'd like to
>>>> run
>>>>>>>> all of our tools and ensure those are working. Another thing we are
>>>>>>> trying
>>>>>>>> is to introduce failures on this cluster when it is under load and
>>>> ensure
>>>>>>>> that there is no data loss.
>>>>>>>>
>>>>>>>> So far, we've been working on stabilizing this cluster and fixing
>>>> bugs.
>>>>>>>> Next week, we will be working on tools and setting up audit so we
>> can
>>>> do
>>>>>>>> some data loss analysis, if any. This will probably take another
>>>> month.
>>>>>>>> After that, I think we should be ready to release a public BETA and
>>>> have
>>>>>>>> our users try it. If we release it sooner than that, I'm not sure it
>>>> will
>>>>>>>> be helpful since tools and simple failure cases might not work as
>>>>>>> expected.
>>>>>>>> As far as formal release goes, I believe end of March or April will