hadoop-common-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Sean Busbey <bus...@cloudera.com>
Subject Re: Jenkins Node Labelling Documentation
Date Tue, 09 Aug 2016 04:01:14 GMT
I'm trying to transition jobs off of the yahoo-not-h2 label, but again
I don't see a single label I can use that covers an appropriate set of
nodes.

Can we expand Hadoop to include H10 and H11? Can we come up with a
label that covers both the H* and the physical ubuntu hosts that have
been puppetized?

On Thu, Aug 4, 2016 at 5:05 PM, Gav <gmcdonald@apache.org> wrote:
>
>
> On Fri, Aug 5, 2016 at 7:52 AM, Sean Busbey <busbey@cloudera.com> wrote:
>>
>> On Thu, Aug 4, 2016 at 4:16 PM, Gav <gmcdonald@apache.org> wrote:
>> >
>> >
>> > On Fri, Aug 5, 2016 at 3:14 AM, Sean Busbey <busbey@cloudera.com> wrote:
>> >>
>> >> > Why? yahoo-not-h2 is really not required since H2 is the same as all
>> >> > the
>> >> > other H* nodes.
>> >>
>> >> The yahoo-not-h2 label exists because the H2 node was misconfigured
>> >> for a long time and would fail builds as a result.
>> >
>> >
>> > Yes I know, but now its not, so is no longer needed.
>> >
>> >>
>> >> What label will
>> >> jobs taht are currently configured to avoid H2 be migrated to? Will
>> >> they be migrated automatically?
>> >
>> >
>> > Currently I'm asking that projects make the move themselves. Most jobs
>> > would
>> > be fine as they have
>> > multiple labels, so just need to drop the yahoo-not-h2 label to give
>> > them
>> > access to H2. If, when I drop the label I
>> > see jobs with it in use, I'll remove it.
>> >
>>
>> I don't see a label I can move to that covers the same machines as the
>> current yahoo-not-h2 nodes and H2. It looks like a union of "hadoop"
>> and "docker" would do it, but "docker" is going away. Also I have to
>> have a single label for use in multi-configuration builds or jenkins
>> will treat the two labels as an axis for test selection rather than as
>> just a restriction for where the jobs can run. I could try to go back
>> to using an expression, but IIRC that gave us things like *s in the
>> path used for tests, which was not great.
>>
>> Can we maybe expand the Hadoop label? (or would "beefy" cover the set?)
>>
>> If the H* nodes are all the same, why do we need the labels HDFS,
>> MapReduce, Pig, Falcon, Tez, and ZooKeeper in addition to the Hadoop
>> label?
>
>
> I was thinking the same yep, those could all go too imho but wanted to
> discuss
> that one seperately.
>
> Gav...
>
>>
>>
>>
>> --
>> busbey
>
>



-- 
busbey

---------------------------------------------------------------------
To unsubscribe, e-mail: common-dev-unsubscribe@hadoop.apache.org
For additional commands, e-mail: common-dev-help@hadoop.apache.org


Mime
View raw message