21:00:23 #startmeeting scientific-sig 21:00:24 Meeting started Tue Jan 19 21:00:23 2021 UTC and is due to finish in 60 minutes. The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:00:25 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 21:00:27 The meeting name has been set to 'scientific_sig' 21:00:36 ahoy 21:01:04 hello Stig :) 21:01:20 Hey Martial! 21:01:23 #chair martial 21:01:24 Current chairs: martial oneswig 21:01:28 What's new? 21:02:57 not much, microstack is fun for testing things :) 21:03:00 This week we are having our twice-yearly company "design summit" - thrashing out lots of planning for R&D directions. An interesting week 21:03:28 oh that is cool 21:04:06 You get to do this stuff with a small firm :-) 21:06:34 welcome julian 21:06:48 Hi martial. Apologies for running late. 21:07:06 Hi julianp 21:07:35 Hello oneswig. 21:08:59 What's on the agenda for tonight? Apologies if it was already covered. 21:09:43 Sorry, I don't think one has been set, unless youhave something up your sleeve martial? :-) 21:09:59 Hi everybody 21:10:12 well there is the ISC email? 21:10:13 Hi jmlowe 21:10:23 also SC21 just opened its CFP 21:10:42 That hasn't gone virtual yet has it? 21:10:52 ISC is 21:10:53 Saw the subjects, didn't read further. What's going on? 21:10:58 SC ... not sure 21:11:01 I'm guessing it's going to be right on the edge of in person 21:11:19 ISC: is virtual but our BoF from last year is on if we want it 21:13:11 that's it :) 21:14:25 I thought ISC didn't accept the BoF submission - did they, and then it got dropped in the confusion of going virtual? I've forgotten.. 21:14:43 okay now I am confused too 21:15:18 I though they offered us to do it remotely last year and then said they will automatically accept it for this year 21:15:19 b1airo would know. 21:15:54 I read "We are aiming to accommodate all the BoFs accepted for ISC 2020 to take place at ISC 2021 Digital. However, if your circumstances have changed and you cannot organize a BoF at ISC 2021 Digital, please let us know early via email." 21:16:16 this seems to say that we are "digital" but we are there? 21:17:39 It does indeed. 21:18:06 so there is that :) 21:18:41 we ought to confirm the original list of participants and confirm everyone can attend again 21:18:59 looking now 21:21:09 Ah, it was the big one: Tim Randles, Martial Michel, Tim Bell, Sadaf Alam, DK Panda, Ildiko Vancsa, Blair & me 21:21:20 That's a lot 21:22:48 OK, got it - details from Blair: The news came through late last week that the ISC organising committee have decided to shelve the BoF programme this year as they think a virtual format will detract from the highly interactive goal of these sessions. But the good news is they think this year's BoF proposals will all be relevant next year too, and so are going to invite us to run it at ISC2021. I 21:22:48 'll assume, unless I hear otherwise, that you are all interested to participate. 21:23:14 That was April last year 21:23:28 well virtual is going to be interesting to say the least but I am game as always 21:26:58 What was the SC news martial? Just that the CFP is open? 21:27:22 yes, pretty much 21:28:06 https://www.irccloud.com/pastebin/racXJSp5/ 21:28:49 jmlowe: doing another workshop at SC ? 21:29:23 probably, we are doing the SuperCompCloud for ISC 21:29:47 Ooh, nice. 21:30:03 jmlowe: will you attempt an HPL number on Jetstream2? 21:30:46 I don't see how we can skip that 21:31:05 I'm going to try for an IO500 submission as well, time permitting 21:31:39 Based on the timing HPL run would probably be submitted for SC'21 not ISC'21 21:31:48 Get on! Is that with CephFS? 21:31:52 yes 21:32:09 nice :) 21:32:37 I'll fire up a native cephfs manila share and see what happens 21:35:15 oops, i got caught in another meeting 21:35:33 It's okay, we gave you all the action items :) 21:35:40 #chair b1airo 21:35:41 Current chairs: b1airo martial oneswig 21:35:41 :-o 21:35:55 morning b1airo 21:36:00 we were talking about ISC and SC 21:36:41 We should discuss future agenda items - would be good to bring in some fresh ideas? 21:39:32 (sorry have to excuse myself for a few minutes) 21:39:36 function as a service everywhere? 21:41:26 Is that project still going? Forgive my ignorance. 21:42:19 which one? the globus thing, funcx? 21:42:41 Not heard of that, was thinking of the OpenStack project 21:44:28 ah right, i can't remember there ever being anything in the tent that really addressed this 21:44:49 oh yeah, Qinling 21:45:47 https://github.com/openstack-archive/qinling - RETIRED 21:46:01 What's the mechanism behind funcx? 21:48:17 it basically allows you to register containers (e.g. Singularity) and entry points as containers, there's an endpoint that you launch on your HPC that then connects to centrally managed Globus services. the API then allows you to run functions (containers) on endpoints (HPCs etc) directly through simple Python 21:50:30 Sounds a bit like Dask without the graph model? 21:51:07 ya not an unreasonable comparison for sure, though aimed at different levels 21:51:59 b1airo: do you have users of funcx? 21:52:29 we have an instrument data processing project in flight at the moment where we are kicking the tyres on it 21:53:35 my understanding from Kyle (Chard) is that it's still more a research project at the moment, but it's leveraging Globus Labs, so that makes it a more reliable prospect 21:54:30 Does it provide anything for data access? Is it all S3 based? 21:54:57 that project is already using Globus and its transfer APIs to do automated data movement, so being able to stay in the same API ecosystem to remotely launch/manage associated processing is nice 21:56:18 Good to know, thanks. 21:57:41 so we have instrument/s at org/uni producing data that needs post-processing. data gets captured into a repo locally, once it lands there either the repo triggers a Globus transfer to NeSI (or Globus just watches a folder on a managed endpoint), data arrives at NeSI, then we use Funcx to kick off the associated processing workloads (which in many cases can be standardised), outputs get Globus'd back to the 21:57:41 original org/uni's Globus endpoint 21:58:02 oneswig: was there a Ceph thing you wanted to discuss? 21:59:56 Not sure. I had some hairy fun upgrading a cluster in place from CentOS7/Docker to CentOS8/Podman. The monitors on the CentOS8 side wouldn't join quorum with the existing monitors. Was that it? 22:00:13 Or was it related to all-flash or all-hdd Ceph configurations? 22:00:38 yeah the Slack convo - will continue it there seeing as appears to be only us here anyway :-) 22:00:48 and we're at time 22:00:54 Time at the bar anyway. 22:01:14 OK, better close the meeting. Thanks all 22:01:17 #endmeeting