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

Switch to Threaded View
Zookeeper >> mail # dev >> Roadmap?

we've done roadmaps in the past, and they don't seem to work out. in
the end it all really depends on the contributions that we get. making
releases feature based tends to encourage the just-in-time
implementation of the required features at the last minute, which
leads to a delayed and unstable release. it might seem like we can at
least suggest direction of work for the next release, but there are
really two key things we need to continually work on: stability and
testing. imo all the rest: scalability, performance, manageability,
features, etc. will come as they are ready and make sense.

perhaps the roadmap should be more time based than feature based: july
1st, we will be doing a release, so in june it will be stability and
test patches that are going in, and then in july 1st we release what
is ready.


On Wed, Jan 4, 2012 at 12:00 PM, Patrick Hunt <[EMAIL PROTECTED]> wrote:
> Our first goal is always reliability, "it goes without saying" which
> is why I didn't highlight it. Every release shoots for the highest
> level of reliability we can provide. I think everyone is in agreement
> on that. If we find a serious issue we take it seriously, _that's_ why
> we had a number of follow-on fix releases so quickly.
> That said, 3.4 is not currently "stable", it's alpha/beta/gamma
> whatever that means these days. It was discussed on list although that
> didn't make it into the announce as clearly as I would have liked.
> On the soapbox side, we rely on everyone to help us guarantee
> reliability of a release. It's not "the commiitters job" or the PMC's
> job, etc... It's the community's job. So if you have issues look to
> yourself first. :-) Find the bugs, report them, help us fix them and
> improve the existing tests -- if you notice that's exactly what we did
> as a result of getting 3.4.0 out there and in more user's hands. If
> you think we should do a better job in testing, well, then add more
> tests (that's what I do).
> Regards,
> Patrick
> On Wed, Jan 4, 2012 at 11:46 AM, Mark Miller <[EMAIL PROTECTED]> wrote:
>> On Wed, Jan 4, 2012 at 2:42 PM, Thomas Koch <[EMAIL PROTECTED]> wrote:
>>> Patrick Hunt:
>>> > It's fine to discuss the roadmap on list. We should only be making
>>> > decisions on list regardless. Last I heard from Mahadev he was
>>> > coordinating the meetup for sometime this month or early next.
>>> So just for the record, I don't expect many to agree:
>>> It's my understanding that ZooKeepers first duty is reliability. The last
>>> major release (3.4.0) however was immediately followed by two bugfix
>>> releases
>>> and jira already shows another 5 blockers and 27 critical bugs.
>>> Maybe it would be a good idea now to make robustness and stability the
>>> first
>>> priority and refuse any new features for the next time.
>>> As for Patrick's idea of scalability I think that there is a lot of
>>> potential
>>> to make ZK scale even more. I've already outlined the idea of an immutable
>>> datatree and the possibility to handle read requests in parallel,
>>> independent
>>> from the write queue. The recently discussed "Disruptor" system[1] is also
>>> worth a look for inspiration. But it would be hazardous if not impossible
>>> to
>>> explore any such path without cleaning up the code first.
>>> [1] http://code.google.com/p/disruptor
>>> Best regards,
>>> Thomas Koch, http://www.koch.ro
>> That release (3.4) was kind of oddly post scripted as an alpha. I didn't
>> catch that at first but did eventually.
>> --
>> - Mark
>> http://www.lucidimagination.com