aurora-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Bill Farner <wfar...@apache.org>
Subject Re: Combining vagrant master machines
Date Tue, 01 Apr 2014 20:01:50 GMT
I'm not convinced we're yet in a position to trigger the complicated class
of bugs you describe, Kevin.  However, the case for being able to exercise
basic functionality is pressing.  I'd rather focus on enabling developers
to do this ASAP, and focus on the more complicated setups later.

-=Bill


On Tue, Apr 1, 2014 at 12:55 PM, Kevin Sweeney <kevints@apache.org> wrote:

> I think keeping a multi-machine setup for testing is important since it's
> much closer to actual deployment (and it'll catch a class of bugs that
> might not show up when everything's on localhost). Also as a technical note
> as far as I know you can't run more than one slave instance per VM when
> you're using the cgroups isolator (due to global kernel state) and the
> multi-slave test case is important as well.
>
> I'd say shoot for a refactoring where you can have 2 setups, one single-VM
> low-footprint one for dev and one multi-VM high-footprint setup for e2e
> testing, aiming for a minimum of code duplication.
>
>
> On Tue, Apr 1, 2014 at 12:16 PM, Jake Farrell <jfarrell@apache.org> wrote:
>
>> Whats the end goal that we want to have with the vagrant image, easier
>> development and testing or as a sample setup for a small cluster. I think
>> that for simplicity keeping it all on one vm for development and initial
>> testing is the best approach and then perhaps for the end to end tests we
>> can use the same chef scripts with a multi-vm.
>>
>> -Jake
>>
>>
>> On Tue, Apr 1, 2014 at 3:06 PM, Bill Farner <wfarner@apache.org> wrote:
>>
>> > Should i take this a step further and use one machine for everything?  I
>> > realize this is less 'pure', but purity doesn't help folks much if they
>> > need to buy a new computer to use it.
>> >
>> > -=Bill
>> >
>> >
>> > On Tue, Apr 1, 2014 at 9:06 AM, Jake Farrell <jfarrell@apache.org>
>> wrote:
>> >
>> >> +1, good call
>> >>
>> >> -Jake
>> >>
>> >>
>> >> On Tue, Apr 1, 2014 at 11:43 AM, Bill Farner <wfarner@apache.org>
>> wrote:
>> >>
>> >> > Does anybody have an issue with me combining the 'master' components
>> in
>> >> our
>> >> > vagrant configuration to one machine?  Currently we have three
>> machines:
>> >> > zookeeper, scheduler, master.  It should be straightforward to
>> combine
>> >> > these into a single machine, at which point 'vagrant up' is much
>> faster
>> >> and
>> >> > can be run on a machine with less RAM.  This is a part of my effort
>> to
>> >> > improve the vagrant setup [1].
>> >> >
>> >> >
>> >> > -=Bill
>> >> >
>> >> >
>> >> > [1] https://issues.apache.org/jira/browse/AURORA-299
>> >> >
>> >>
>> >
>> >
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message