11:00:13 #startmeeting scientific-sig 11:00:13 Meeting started Wed Jul 14 11:00:13 2021 UTC and is due to finish in 60 minutes. The chair is oneswig. Information about MeetBot at http://wiki.debian.org/MeetBot. 11:00:13 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 11:00:13 The meeting name has been set to 'scientific_sig' 11:00:21 greetings 11:01:23 Having fun over here reclaiming disk space from a docker registry... 11:05:06 hi, didn't know if there was a meeting today :) I am reviewing network change controls 11:05:50 Hi dh3 there isn't an agenda, as usual I have been distracted on other matters. 11:05:57 o/ 11:06:03 OpenStack-related network changes? 11:06:08 hi belmoreira, welcome 11:06:29 hi oneswig 11:06:36 only tangentially (more network switches to support more sequencing machines... the data will go through OpenStack at some point) 11:07:07 One interesting development this week has been the arrival of OFED 5.4 for HPC networking. It doesn't appear to work out of the box for us. 11:07:36 dh3: more sequencing machines? The data deluge is upon us 11:07:49 always /o\ 11:08:18 I suspect belmoreira is looking at similar issues for the high-luminosity upgrade? 11:08:34 I had not touched OFED in years until I had to do a kernel upgrade on an elderly Lustre system a few weeks ago 11:09:02 dh3: sounds potentially traumatic 11:09:12 :) 11:09:38 I would like to bring something to the SIG 11:09:48 belmoreira: fire away! 11:09:56 One of the discussions in the TC is the goals for the Y release. A proposal is the "operators pain points" in the different projects. 11:10:15 One of the problems is how to collect those pain points. 11:10:49 Pain points that the TC can do something about, specifically... 11:10:56 It would be great if the scientific SIG community has some feedback 11:11:15 I think I remember questions in the OpenStack survey but maybe those were project-specific 11:11:18 belmoreira: sounds like a good fit 11:12:03 operators pain points that the OpenStack projects can work on. The problem is that dev/operators can have a different understanding of pain points 11:12:31 that's way it would be great to have operators feedback 11:13:02 belmoreira: I wonder if the TC can do anything on settling volatility in the CentOS roadmap 11:13:10 yes, OpenStack survey questions are one of the information sources. 11:13:40 oneswig ahah 11:14:51 here's the discussion in the mail list: http://lists.openstack.org/pipermail/openstack-discuss/2021-June/023352.html 11:16:34 I think I will have one specific pain point later this year: in our upgrade we will be live-migrating many instances around for as close to a seamless upgrade as possible. However hardware maintenance, repairs, replacement etc have left some hypervisors subtly different (microcode version) which breaks live migration. Not sure what could be done to suggest/recommend CPU feature settings to hide this sort of thing from day 1 11:17:18 dh3 I had that problem as well 11:17:36 dh3: to compound on that, I think CentOS 8.3 revoked some flags (for TSX) - for good reasons. 11:17:38 we mitigate it using now a custom CPU features 11:18:12 I guess you revisit the CPU feature set every time you have a significant hypervisor CPU change/new hardware? 11:19:48 oneswig: I don't *think* that will bite us (I don't see TSX in a random VM I just checked) but good to know about, thanks 11:22:39 dh3, true those need to be revisited when the hardware is replaced 11:24:06 belmoreira: what's the best way to contribute feedback to each project? 11:25:05 Should we add them here - https://etherpad.opendev.org/p/y-series-goals - or suggest them for review somewhere else first? 11:25:51 oneswig that's a great question. That's not defined yet. I just wanted to give you a first heads up. 11:26:44 It's a great opportunity for feedback and I hope the hard work will be on reducing a long list to a short one. 11:27:33 morning 11:27:50 Hi martial, welcome 11:27:54 #chair martial 11:27:54 Current chairs: martial oneswig 11:27:58 (to be honest we have not had a huge amount of operator pain recently... clearly it's time for us to start using more projects...) 11:28:11 We were just discussing operator pain points to focus on for the Y cycle 11:28:32 hi martial 11:29:07 understood, hello all 11:29:14 unfortunately I need to leave for another meeting. I was good talking with you. 11:29:28 likewise belmoreira, thanks! 11:29:29 thanks, bye 11:29:31 I will give you more info when available 11:29:38 thanks 11:31:01 okay will read the logs 11:32:18 The summary is that we should canvas the SIG for pain points with specific projects. Just annoying dumb stuff that nobody ever fixed, but should have... 11:35:35 I can think of one we hit the other day - I don't think Glance enforces quotas for image usage. Consistent implementation of quotas, roles and policies would be helpful. 11:36:44 makes sense 11:37:08 martial: shall we add it to the agenda for next week to produce a shortlist? 11:37:38 maybe start a slack conversation to gather items? 11:38:03 Sounds good to me. 11:39:52 I can start that one martial. Is there anything else to cover today? (I'm eyeing an opportunity for lunch before the afternoon kicks off) 11:40:37 nothing from my end 11:40:51 go enjoy your lunch :) 11:42:43 Great, thanks all, see you next time 11:43:26 Very late hello/goodbye. 11:43:59 hi verdurin 11:44:00 we were about to close the meeting 11:44:32 Yes, please go ahead. 11:44:33 I think it is the two of us left :) 11:44:42 #end-meeting 11:44:58 mmh weird 11:45:21 #endmeeting