11:01:11 <oneswig> #startmeeting scientific-wg
11:01:12 <openstack> Meeting started Wed Aug 30 11:01:11 2017 UTC and is due to finish in 60 minutes.  The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot.
11:01:14 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
11:01:16 <openstack> The meeting name has been set to 'scientific_wg'
11:01:21 <oneswig> Hello all
11:01:32 <enolfc> Hi
11:01:34 <verdurin> Afternoon, oneswig
11:01:37 <oneswig> #link Agenda for today https://wiki.openstack.org/wiki/Scientific_working_group#IRC_Meeting_August_30th_2017
11:01:49 <oneswig> Top of the afternoon to you all :-)
11:02:00 <oneswig> #chair martial
11:02:01 <openstack> Current chairs: martial oneswig
11:02:07 <oneswig> #chair b1airo
11:02:08 <openstack> Current chairs: b1airo martial oneswig
11:02:11 <b1airo> o/
11:02:27 <oneswig> The three horsemen of the apocalypse, reunited at last...
11:02:44 <oneswig> Hi b1airo
11:02:49 <oneswig> In other words
11:03:03 <b1airo> how's things? feeling refreshed oneswig, one week wouldn't cut it for me!
11:03:32 <b1airo> honestly still fantasizing about being on the beach
11:03:32 <oneswig> I was fully recharged for at least a day...
11:03:36 <martial> glad you gents had "hopefully relaxing" vacations :)
11:03:43 <b1airo> hah
11:04:02 <oneswig> It's almost back to school time, alas
11:04:17 <oneswig> OK lets get going - just the second edition to discuss on the agenda today
11:04:30 <oneswig> #topic second edition of the Scientific OpenStack book
11:04:48 <oneswig> OK, I read most of the first edition again, a real page turner
11:05:03 <martial> oneswig: maybe oenstack days london if that is still something to finalize?
11:05:07 <b1airo> it was certainly gripping stuff
11:05:49 <oneswig> Kathy Cacciatore and Chris Hoge from the Foundation are keen to produce a revised edition, as you'll probably know by now
11:06:28 <oneswig> Enol has stepped up to lead the creation of a sixth chapter on federation - correct enolfc?
11:06:55 <oneswig> martial: OS days London - good point
11:07:32 <oneswig> There are three major tasks to cover in the second edition:
11:07:37 <enolfc> oneswig: yes ,starting to work on this
11:07:41 <oneswig> 1) refresh the good bits
11:07:50 <oneswig> 2) cut out and replace the bits that have 'gone off'
11:07:58 <oneswig> 3) add the federation chapter
11:08:17 <oneswig> enolfc: that's great to hear.
11:08:54 <oneswig> I will contact the contributors to the original book and ask for updates
11:09:13 <priteau> I hope the Chameleon chapter is considered one of the good bits, as I have started to refresh it ;-)
11:09:46 <oneswig> The book was assembled in a set of google docs, and then transcribed to markdown (in one direction) and something more graphic-design oriented (for print)
11:09:48 <b1airo> hi priteau!
11:10:02 <oneswig> Thanks priteau, that's one less email to send :-)
11:10:12 <priteau> oneswig: We'll work directly on the scientific-wg repo this time?
11:10:36 <oneswig> priteau: that's something I wanted to ask people about.
11:10:41 <verdurin> oneswig: on Github, it says it's in RST
11:10:58 <priteau> Yes, it's RST
11:11:14 <oneswig> verdurin: markdown/rst, whatever, I get them mixed up, apologies
11:11:32 <priteau> Markup format overflow
11:12:44 <priteau> oneswig: I am used to gerrit so it's not a problem for me. Maybe more complex for newcomers?
11:12:45 <oneswig> The question I had was, do people prefer to use the gerrit process, or suggestion mode on a google doc?
11:12:51 <b1airo> i am yet to see a Turing complete markdown...
11:12:53 <priteau> Especially with the strict deadline
11:13:18 <enolfc> I'm fine with sending reviews to gerrit also
11:13:29 <b1airo> likewise
11:14:11 <oneswig> OK, that sounds fine by me.
11:14:17 <daveholland> haven't used gerrit but happy to learn (assuming it's like github pull request?)
11:14:23 <b1airo> if it will ultimately be typeset and published from there then it makes sense to stay there
11:14:59 <verdurin> I'm rusty with Gerrit, happy to go with the consensus
11:15:03 <b1airo> but perhaps if it suits anyone in particular we could let them edit in Docs and do the reviews for them (rather than miss out on content)
11:15:43 <oneswig> We have the advantage that all changes are disjoint from one another.
11:15:59 <oneswig> Only one person is responsible for each section of text.
11:16:34 <priteau> oneswig: Does it mean that all authors need to sign the Contributor License Agreement, and potentially their employer need to sign the Corporate Contributor License Agreement too?
11:16:36 <oneswig> So we can do either - but the final handover to Kathy is a google doc.  I can transfer patchsets from gerrit.
11:17:33 <oneswig> priteau: I don't know about that.  Is that applied to all OpenStack repos?
11:17:57 <oneswig> It didn't come up last time because it was done in Google doc - outside of the OpenStack repos.  The document is creative commons licensed.
11:18:32 <b1airo> hmm, sounds like a question for governance peeps
11:18:58 <oneswig> Do we have a case where that might be contentious?
11:19:08 <b1airo> i recall some changes happening around this recently so that it was less of a blanket requirement, but i wouldn't be surprised if gerrit access relied on it
11:19:13 <priteau> oneswig: I had a collegue start contributing to an OpenStack project recently, and he couldn't push to Gerrit until he had signed the CLA
11:19:29 <priteau> b1airo: I believe the change is that you don't need to join the OpenStack foundation anymore
11:19:52 <oneswig> The simple way around this is to migrate editing to google docs.
11:19:52 <b1airo> ah yep, thanks priteau
11:20:27 <b1airo> (for anyone not willing or able to sign?)
11:20:29 <priteau> oneswig: Not sure if it can be contentious, but potentially lengthy if one has to deal with their legal department
11:20:50 <oneswig> priteau: right - I once worked in such a place.
11:21:18 <priteau> UChicago was quite efficient, I was impressed :-)
11:21:47 <oneswig> OK, it's not insurmountable.  I will provide both options.
11:22:06 <b1airo> depends very much on the objectives of the org.!
11:22:07 <oneswig> enolfc: are you working on the RST in git?
11:23:01 <enolfc> oneswig: yes, but don't mind moving to google doc
11:23:10 <martial> b1airo: as much as people are willing to sign, sometimes their organization make it hard for them
11:24:02 <oneswig> enolfc: entirely up to you.  Whichever way is simpler for managing and combining contributions.
11:24:24 <b1airo> yep, that's what i meant - some legal departments are all about protecting company IP at all costs and no matter how measly
11:24:52 <oneswig> enolfc: the final output should be a google doc, that's all we require.  The process is not important
11:25:11 <priteau> So should we leave it to each author to choose between gerrit and GDocs?
11:26:01 <b1airo> sounded like that was the outcome
11:26:20 <oneswig> priteau: either works for me, once editing is complete it's easy to combine into the final doc
11:26:56 <b1airo> so oneswig, will you ping everyone with those details and a drop-dead date on edits, with links to the current/last version of the docs?
11:27:12 <oneswig> b1airo: I will do that
11:27:25 <martial> how are we organizing the editor/reviewer jobs?
11:27:39 <verdurin> martial: my offer to be a reviewer stands
11:27:52 <martial> verdurin: thank you
11:27:57 <b1airo> likewise i am happy to review one or two others
11:28:23 <oneswig> verdurin: can you (and anyone else able to review) add yourself to https://etherpad.openstack.org/p/scientific-computing-book-update-2017
11:28:48 <martial> if we are going to use Google docs, we can ask people to male "suggestions" and "comments" on the individual documents
11:29:15 <oneswig> martial: agreed
11:29:20 <martial> I would keep each article separate until it is time to merge, because people editing at the same time create chaos ;)
11:29:33 <priteau> My availability for September will be patchy, so I can't commit to review, but if available I will try to look at what's been proposed
11:29:47 <enolfc> martial: +1
11:30:13 <oneswig> priteau: if you can refresh the chameleon case study, that would be a great help.  Reviewing elsewhere is a bonus
11:30:43 <oneswig> daveholland: Are you and James still able / interested in contributing a case study on Lustre at Sanger?
11:30:46 <priteau> The Chameleon case study should be no problem
11:30:57 <zioproto> I probably will not have time to participate into this, but I think google docs is much better for this kind of review/edits. We are not really reviewing code here, where Gerrit is perfect. To work on text documents Google Docs works just fine.
11:31:15 <daveholland> oneswig: yes, we spoke about it this morning. Do you have any guidance on e.g. number of words, audience knowledge level...?
11:31:57 <oneswig> zioproto: I tend to agree, but we can offer both options
11:31:59 <martial> daveholland + oneswig: I agree, we ought to place a number of pages limit and style
11:32:42 <martial> what we know:  Selected 8"H x 10"W.
11:32:51 <oneswig> daveholland: only the first edition as a guide.  Roughly, it works out as a page, including a picture and a quotation.  So about half a page of text, including some particular details
11:33:12 <b1airo> daveholland, it's mostly digital so length is not a big deal but take a look at the existing book - the sections are relatively short
11:33:32 <daveholland> oneswig: OK that's less than we expected (depending how much explanation of Lustre routers we need to put it)
11:33:34 <martial> oneswig: 10 pages limit, single column
11:34:40 <oneswig> daveholland: in terms of the level of technical detail, assume you're writing for somebody attending Supercomputing who knows about OpenStack.  Technical architect types
11:34:54 <b1airo> should we review the list of V2 suggestions that Kathy provided?
11:35:25 <oneswig> martial: thanks - that's for a chapter, IIRC?  And the pages are google doc pages
11:35:43 <oneswig> b1airo: mostly they were typos etc?
11:37:52 <oneswig> We should look at case studies for federation - enolfc have you thoughts on how to get diverse coverage of case studies?
11:38:43 <oneswig> And we also need a new case study for bare metal infrastructure management
11:39:29 <martial> enolfc: are you working with Khalil on content as well?
11:39:53 <enolfc> oneswig: not yet
11:40:28 <enolfc> I need to contact Tim for the possible CERN input
11:40:31 <martial> stig: do we have a directory in google drive for people to populate
11:40:43 <oneswig> I'd suggest we include three, chosen to take different approaches and (ideally) be geographically spread
11:41:03 <martial> with their content and then we can act as editors as well as writers to integrate content once it is vouched for?
11:41:05 <oneswig> martial: There was one, I'll check
11:41:21 <b1airo> enolfc, i would imagine Wilfred could pull together some content about how the Nectar federation works (and how we see it evolving)
11:41:22 <martial> if there was one, it needs to be linked into the etherpad
11:41:50 <enolfc> blairo, oneswig: so we can have Nectar, CERN, and EGI
11:42:07 <oneswig> enolfc: that would work - provided they don't all use the same solution!
11:42:13 <martial> make it "v2" or something, we let people put new content there and we integrate when we are ready
11:42:24 <oneswig> martial: good point, will do
11:42:33 <enolfc> something from the US would be nice
11:42:37 <martial> thanks Stig
11:42:48 <b1airo> along the lines of oneswig's question re. the same solution - is CERN's cloud federated heavily in any particular dimension?
11:43:01 <oneswig> enolfc: agreed.
11:43:12 <martial> enolfc: Mike Lowe, Robert Budden and Time Randles were talking about adding content
11:43:20 <b1airo> they are big users of Cell's, but I am not sure if that is Cell's across multiple orgs?
11:43:32 <oneswig> b1airo: they use federated authentication for access to CERN tier-0 resources IIRC
11:43:46 <oneswig> so not multi-site so much for the infras
11:44:16 <oneswig> apart from Hungary I guess, but it's still logically the same 'site'
11:44:33 <b1airo> oneswig, but i presume much like our Australian Access Federation that the Federation part is actually a separate and previously established service?
11:45:17 <b1airo> perhaps it is more deeply integrated into the cloud access?
11:45:31 <b1airo> e.g. SSO?
11:46:09 <oneswig> I don't think the federated auth pre-dates OpenStack, I am not sure.  I have notes from a workshop I attended (with zioproto - perhaps he remembers better...)
11:46:24 <oneswig> and enolfc - weren't you atteding remotely?
11:46:58 <b1airo> no sign of of noggin or belmiro unfortunately
11:47:15 <b1airo> anyway, we shall see i guess - i am certainly curious
11:47:15 <zioproto> oneswig: in the NREN context federated out came much earlier than Openstack and it is used for many other services
11:47:32 <zioproto> s/federated out/federated auth/
11:47:33 <oneswig> EduGAIN?
11:47:45 <zioproto> oneswig: yes
11:48:02 <b1airo> Shibboleth has been around quite a while now, for example
11:48:03 <martial> I understand we have two more topics to cover in the next few minutes (OpenStack Days London + SC17 notes), so I am going to see if we can move on to the next topic at 50
11:49:27 <oneswig> OK - thanks martial.  enolfc - I'm sure you can find diversity in the case studies?  There are others that might also work - SNIC in Sweden, the Finnish one, for example
11:49:49 <oneswig> We just like to see good coverage.
11:50:13 <oneswig> OK, final comments?
11:50:20 <oneswig> lets move on
11:50:21 <martial> if you have more comments on the book chapter, please refer to the ehterpad at
11:50:23 <martial> #link https://etherpad.openstack.org/p/scientific-computing-book-update-2017
11:50:35 <oneswig> #topic AOB
11:50:43 <oneswig> London OpenStack
11:51:17 <oneswig> coming up on the September 26th and we have a WG session, arranged to be lightning talks
11:51:58 <oneswig> For those that will be going - lets have a short talk from you on something interesting going on where you are!
11:52:17 <enolfc_> had a network glitch... not sure my last message was sent, but I will contact people and let you know
11:52:27 <oneswig> OK, thanks enolfc_
11:53:04 <oneswig> I will ask around for an "incentive" for good talks
11:53:20 <verdurin> oneswig: I could talk about what I plan to do where I will be next? Still sketchy, though...
11:54:09 <oneswig> verdurin: Your last talk was largely in the past tense, why not?
11:55:11 <verdurin> Hopefully there will be others talking in a more stable situation
11:55:23 <oneswig> The other item was SC - martial would you like to update on last week?
11:55:33 <martial> stig:sure
11:56:03 <b1airo> ooh goodie - i haven't read last week's logs yet
11:56:08 <martial> basically Mike Lowe mentioned that "<jmlowe> I have the green light for a round of lightning talks in the IU booth just like last year"
11:56:32 <martial> "<jmlowe> I'll have 1 to 1.5 hours, so divide by volunteers and..."
11:56:46 <b1airo> ah dang, i think i knew that already *deflated* :-)
11:57:08 <martial> #link https://etherpad.openstack.org/p/SWG-SC17
11:57:34 <martial> that is the working document, we will need to coordinate with Mike next week, but is gauging interest
11:57:55 <oneswig> b1airo: are you going?  Given you're barely travelling the week before...
11:58:00 <b1airo> maybe by then i will have a talk on "what workloads not to virtualise" :-)
11:58:17 <oneswig> b1airo: ooh, interesting, based on what?
11:58:19 <b1airo> yes i'll be there oneswig
11:58:45 <verdurin> Reasonable chance I'll be there
11:58:56 <b1airo> based on when we see THP defrag going mental
11:59:20 <oneswig> b1airo: ah, that
11:59:30 <oneswig> Final comments anyone?
11:59:48 <oneswig> b1airo: does it do that for CryoEM?
11:59:54 <zioproto> it is the same dates as Sydney ?
11:59:59 <oneswig> week after
12:00:07 <b1airo> yeah, my analysis of that stalled, but also the incidence of *known* occurrences has dropped off sharply since we started aggressively dropping caches
12:00:21 <martial> Nov 12-17 I think
12:00:26 <verdurin> Yes
12:00:37 <martial> #link http://sc17.supercomputing.org/
12:00:38 <b1airo> but i will go back and get more data
12:00:44 <oneswig> OK, we are out of time, alas.
12:00:56 <b1airo> really want to know why having used page cache makes the VMM stupid
12:01:09 <oneswig> Thanks everyone, got to close
12:01:11 <verdurin> Bye
12:01:13 <oneswig> #endmeeting