aurora-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From ASF IRC Bot <asf...@urd.zones.apache.org>
Subject Summary of IRC Meeting in #aurora
Date Mon, 28 Apr 2014 18:53:57 GMT
Summary of IRC Meeting in #aurora at Mon Apr 28 18:04:01 2014:

Attendees: hsaputra, mohitsoni, wickman, jfarrell, mchucarroll, wfarner, jaybuff, dlester,
kts_

- Preface
- first release
  - Action: jfarrell verify py headers
  - Action: wickman py style checker to automate license header check for this for future
releases
  - Action: dlester add README update/cleanup to AURORA-147
- MesosCon
  - Action: wfarner aurora abstract to wiki for MesosCon
- release format
  - Action: kevints+dlester installation / deployment documentation


IRC log follows:

## Preface ##
[Mon Apr 28 18:04:36 2014] <wickman>: correct.  the scheduler holds onto offers for
a period of time.
[Mon Apr 28 18:04:45 2014] <wickman>: this is purely to be more responsive if a new
job comes in.
[Mon Apr 28 18:05:09 2014] <mohitsoni>: Okay, makes sense. The reason I am asking this
question is because in our internal aurora/mesos installation, we observed that in cluster's
ideal state (when there's no active task), mesos dashboard shows that Aurora's CPU share is
100%.
[Mon Apr 28 18:05:33 2014] <jfarrell>: afternoon everyone, lets start the meeting off
with a roll call of any committers here
[Mon Apr 28 18:05:41 2014] <wickman>: wfarner can probably speak more to the long term
there.
[Mon Apr 28 18:05:41 2014] <kts_>: present
[Mon Apr 28 18:05:47 2014] <dlester>: present
[Mon Apr 28 18:05:50 2014] <wickman>: jfarrell: aye aye cap'n
[Mon Apr 28 18:06:09 2014] <mchucarroll>: I’m here.
[Mon Apr 28 18:07:03 2014] <jfarrell>: right now the primary topic for this weeks meeting
will be around the first release
## first release ##
[Mon Apr 28 18:07:56 2014] <jfarrell>: AURORA-147 sub tickets have comments on each
one and almost everything is in review right now
[Mon Apr 28 18:08:50 2014] <jfarrell>: I've updated the LICENSE file this morning based
on the js changes done earlier in the week and we need to go through one last time and verify
all the headers in the py side
[Mon Apr 28 18:08:58 2014] <jfarrell>: #action jfarrell verify py headers
[Mon Apr 28 18:09:41 2014] <jfarrell>: #action wickman py style checker to automate
license header check for this for future releases
[Mon Apr 28 18:10:16 2014] <jfarrell>: kts will be acting as the first RM
[Mon Apr 28 18:10:21 2014] <wickman>: sorry about not getting to it last week.  i'm
oncall this week so i might have a hard time getting to it this week as well, but next week
is a possibility.
[Mon Apr 28 18:10:23 2014] <wfarner>: late, but present
[Mon Apr 28 18:11:06 2014] <dlester>: Would love to get in a quick review before the
first release to cleanup the README. what we have right now is fairly barebones
[Mon Apr 28 18:11:48 2014] <jfarrell>: #action dlester add README update/cleanup to
AURORA-147
[Mon Apr 28 18:13:48 2014] <jfarrell>: i've been testing the release* scripts as dry-runs,
any objections to me performing a full end-end and then deleting everything post test so we
have full verification prior to the first release
[Mon Apr 28 18:14:42 2014] <kts_>: jfarrell: +1
[Mon Apr 28 18:14:52 2014] <jfarrell>: outside of that i think we are looking pretty
good for the release candidate to occur this week
[Mon Apr 28 18:14:57 2014] <wfarner>: +1
[Mon Apr 28 18:15:27 2014] <jfarrell>: if there are any open tickets not tagged or completed
with 0.5.0 please get them tagged today
[Mon Apr 28 18:16:08 2014] <jfarrell>: anyone against saying tomorrow is a cleanup day
and wed we cut the rc
[Mon Apr 28 18:16:17 2014] <wfarner>: sounds great, +1
[Mon Apr 28 18:16:20 2014] <wickman>: +1
[Mon Apr 28 18:16:22 2014] <dlester>: +1
[Mon Apr 28 18:16:34 2014] <kts_>: +1
[Mon Apr 28 18:17:10 2014] <dlester>: kts_ will be tagging the release, correct?
[Mon Apr 28 18:17:44 2014] <kts_>: that's correct, but the bulk of that operation will
be running jfarrell's scripts
[Mon Apr 28 18:17:47 2014] <jfarrell>: kts will be running the scripts, they create
the tags and auto generate the email
[Mon Apr 28 18:18:09 2014] <kts_>: if all goes well this is just a ceremonial button-pressing
role
[Mon Apr 28 18:18:17 2014] <jfarrell>: release-candidate script, send generated email,
wait for vote to close, release script
[Mon Apr 28 18:18:27 2014] <jfarrell>: send final email
[Mon Apr 28 18:18:46 2014] <kts_>: need to step out to take a call, sorry folks
[Mon Apr 28 18:19:36 2014] <jfarrell>: right now we also have all the dev test tags,
objections to dropping those tomorrow before the rc gets created
[Mon Apr 28 18:20:31 2014] <jfarrell>: and do we want to drop all branches and only
have branches for the rc/releases as well?
[Mon Apr 28 18:22:02 2014] <wickman>: that seems reasonable to me.  we probably should
keep branches on personal github forks anyway.  as for tags, we can always archive those on
personal forks as well.
[Mon Apr 28 18:23:07 2014] <wfarner>: +1 to killing dev tags on the canonical repo
[Mon Apr 28 18:23:48 2014] <jfarrell>: if there are no objections then i'll clean those
up when running the end-end tests
[Mon Apr 28 18:24:15 2014] <jfarrell>: think that about covers everything for the first
release candidate then
## MesosCon ##
[Mon Apr 28 18:25:45 2014] <dlester>: Would definitely love to see an Aurora talk @
#MesosCon if possible
[Mon Apr 28 18:25:59 2014] <wfarner>: MesosCon CFP is underway, closes May 16.  Within
twitter, we're thinking of proposing 2-3 Aurora talks
[Mon Apr 28 18:26:07 2014] <dlester>: the CFP is online: http://events.linuxfoundation.org/events/mesoscon/program/cfp
[Mon Apr 28 18:26:38 2014] <wfarner>: My plan is to come up with abstracts and email
dev@ next week to solicit feedback on content of the talks
[Mon Apr 28 18:26:58 2014] <dlester>: the idea of an Aurora hackathon to coincide with
the event is an interesting one
[Mon Apr 28 18:27:16 2014] <dlester>: (full-disclosure, I am MesosCon program chair)
[Mon Apr 28 18:27:26 2014] <wfarner>: yes, something more hands-on would be fantastic
[Mon Apr 28 18:27:43 2014] <jfarrell>: +1, wiki would be good place to put the abstract
up at and iterate on
[Mon Apr 28 18:28:07 2014] <wfarner>: jfarrell: good point, i'll take that approach
[Mon Apr 28 18:28:25 2014] <mchucarroll>: wfarner: I don’t know if this is on your
list, but I’d kindof like to try for a talk on client2.
[Mon Apr 28 18:28:32 2014] <wfarner>: in the meantime, i welcome emails to dev@ with
pre-emptive requests for talk content
[Mon Apr 28 18:28:49 2014] <wfarner>: mchucarroll: awesome!  we can propose 3-4 talks,
then :-)
[Mon Apr 28 18:29:04 2014] <wfarner>: dlester: do you know roughly how many talk slots
there are?
[Mon Apr 28 18:30:24 2014] <dlester>: the current program has a morning workshop, 2
keynote, and slots for 7-8 talks overall -- single track
[Mon Apr 28 18:31:25 2014] <wfarner>: aha, so we'll likely want to collapse some.  how
long are the slots?
[Mon Apr 28 18:31:41 2014] <dlester>: 20 minute afternoon talks
[Mon Apr 28 18:32:25 2014] <dlester>: If there's overwhelming demand/interest, we may
have to switch things up -- possibly an evening event as well. largely dependent on sponsorship
and demand -- I expect that will be more-clear in the next week or so
[Mon Apr 28 18:33:07 2014] <wfarner>: thanks, dlester.  please let us know if things
change
[Mon Apr 28 18:33:42 2014] <wfarner>: stepping out for a meeting, -=wfarner
[Mon Apr 28 18:33:45 2014] <jfarrell>: #action wfarner aurora abstract to wiki for MesosCon
[Mon Apr 28 18:33:47 2014] <jfarrell>: anything else on MesosCon?
[Mon Apr 28 18:34:16 2014] <dlester>: Absolutely, I can report back here. I'm also posting
weekly program committee notes to the dev@mesos.apache list if anyone wants more-details
[Mon Apr 28 18:34:21 2014] <dlester>: I think that's it
[Mon Apr 28 18:34:38 2014] <jfarrell>: any open items anyone would like to bring up?
[Mon Apr 28 18:34:55 2014] <kts_>: jfarrell: i'd like to briefly discuss release format
## release format ##
[Mon Apr 28 18:36:12 2014] <kts_>: so it looks like we'll be publishing a straight tarball
of the source repo on our site for the release
[Mon Apr 28 18:36:41 2014] <jfarrell>: the first release will be a source dist and will
be on dist.a.o which we can link on the site
[Mon Apr 28 18:37:09 2014] <kts_>: we have the ability to publish the aurora schduler
jar to maven and the executor sdists to pypi as well
[Mon Apr 28 18:37:09 2014] <jfarrell>: dist.a.o is what gets replicated to all the mirrors
[Mon Apr 28 18:38:08 2014] <wickman>: the problem with pypi source distributions is
that they will have a requirement on something like 'mesos==0.18.0' and no corresponding packages
there.
[Mon Apr 28 18:38:47 2014] <jfarrell>: there is an open ticket for that in mesos (assigned
to me) was going to look into that after getting our first release out
[Mon Apr 28 18:39:36 2014] <hsaputra>: @jfarrell is there already svnpubsub for Aurora
distribution?
[Mon Apr 28 18:39:37 2014] <kts_>: i'd also like to consider providing a binary distribution
of the scheduler - something made with ./gradlew distTar
[Mon Apr 28 18:40:04 2014] <jfarrell>: hsaputra: yes, i set that up last week. KEYS
files are in place and contain kts_, wfarner and my keys
[Mon Apr 28 18:40:12 2014] <hsaputra>: COOL! Thanks
[Mon Apr 28 18:40:28 2014] <jfarrell>: kts_: that will require a separate LICENSE file
and will take a fair amount of time to put together
[Mon Apr 28 18:40:29 2014] <hsaputra>: I saw you closed the KEYS bug, thanks much
[Mon Apr 28 18:40:54 2014] <kts_>: that's my understanding as well, so before we take
on that work I wanted to see if there was community interest
[Mon Apr 28 18:41:29 2014] <wickman>: how does a published jar work vis-a-vis libmesos?
[Mon Apr 28 18:41:46 2014] <kts_>: there's nothing in jar metadata that lets you specify
that you have a native dep
[Mon Apr 28 18:41:58 2014] <kts_>: the mesos jar is in maven
[Mon Apr 28 18:42:18 2014] <kts_>: and it works fine as long as you have libmesos.so
on your library path
[Mon Apr 28 18:42:39 2014] <wickman>: ah, ok.
[Mon Apr 28 18:42:58 2014] <kts_>: the aurora unit tests actually depend on it and run
without needing a native dep
[Mon Apr 28 18:43:03 2014] <jfarrell>: I would like to do the same for the python package
for mesos as well
[Mon Apr 28 18:44:16 2014] <jaybuff>:  can't mesos just publish meso-python-lib to pypi
and then the executor dep on that?
[Mon Apr 28 18:44:29 2014] <jfarrell>: its system dependent
[Mon Apr 28 18:44:30 2014] <jaybuff>: where mesos-python-lib is just the code that is
generated from swig
[Mon Apr 28 18:44:52 2014] <jaybuff>: lots of python packages have native code... like
numpy
[Mon Apr 28 18:45:06 2014] <jfarrell>: i think that it is worthwhile to make the install
as easy as possible, but we should do it in steps
[Mon Apr 28 18:45:15 2014] <jaybuff>: *nod*
[Mon Apr 28 18:45:50 2014] <wickman>: zc-zookeeper-static is a good example where somebody
took the python subdir of a C-based project and inverted it into a pypi based distribution
[Mon Apr 28 18:45:53 2014] <wickman>: but it's definitely nontrivial.
[Mon Apr 28 18:46:10 2014] <kts_>: so is the consensus that since people will have to
do things a little differently to install mesos, we should be fine with requiring that for
aurora as well?
[Mon Apr 28 18:47:41 2014] <kts_>: at least for now
[Mon Apr 28 18:47:49 2014] <jfarrell>: i think thats fine as long as we document it
for this first release
[Mon Apr 28 18:48:05 2014] <jfarrell>: 1st release source release, 2nd release add the
dist targets (deb, rpm?) and keep working to make it as easy as possible for install from
there
[Mon Apr 28 18:48:53 2014] <dlester>: +1
[Mon Apr 28 18:49:21 2014] <kts_>: +1, I agree we won't get it perfect on release #1
[Mon Apr 28 18:49:46 2014] <kts_>: #action kevints+dlester installation / deployment
documentation
[Mon Apr 28 18:49:50 2014] <wickman>: i've started on pure python impls for both libprocess
and the mesos framework api, but these are purely tinkerings / side projects, but could definitely
lower the barrier for entry in the future for a pip installed aurora executor
[Mon Apr 28 18:51:10 2014] <jfarrell>: MESOS-684 and MESOS-946 are the two tickets tracking
the pypi updates for mesos
[Mon Apr 28 18:51:46 2014] <jfarrell>: anything else anyone would like to cover?
[Mon Apr 28 18:52:43 2014] <kts_>: I think that's it
[Mon Apr 28 18:52:55 2014] <jfarrell>: thanks everyone, lets get anything remaining
tagged with 0.5.0 in jira and get it ready for wed
[Mon Apr 28 18:52:57 2014] <wickman>: thanks, linked MESOS-946 to MESOS-857
[Mon Apr 28 18:53:21 2014] <jfarrell>: any blockers please bring up here in channel
or on the dev@ list otherwise the plan will be for rc wed
[Mon Apr 28 18:53:33 2014] <dlester>: Thanks!
[Mon Apr 28 18:53:36 2014] <jfarrell>: ASFBot: meeting stop dev@aurora.incubator.apache.org


Meeting ended at Mon Apr 28 18:53:36 2014

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