09:03:07 #startmeeting scientific-wg 09:03:08 Meeting started Wed Mar 29 09:03:07 2017 UTC and is due to finish in 60 minutes. The chair is b1airo. Information about MeetBot at http://wiki.debian.org/MeetBot. 09:03:09 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 09:03:11 The meeting name has been set to 'scientific_wg' 09:03:22 morning :) 09:03:30 hi dav 09:03:39 (sorry - missed my autocomplete there) 09:04:03 no "edit one line back" in irc sadly 09:04:23 GMT/BST confused me, I was here an hour ago 09:04:50 oh dear! hope you can hang around for a while longer? 09:05:16 pretty sure we must be about to end daylight savings here too so this swings back to 7pm for me 09:05:50 sure, anything to escape a conversation about per-tenant provider networks, Ansible and policy.json :) 09:07:05 oh come on, how could openstack ever be tedious! 09:07:38 Is tenant networking a bugbear for you at Sanger currently? 09:08:30 we're looking at providing Lustre to tenants using secure Lustre, subnet mounts, (private) provider networks and Lustre routers. Lustre is excitingly complicated, so is OpenStack, let's borg them together... 09:08:55 hi simon-AS559 09:08:59 Hi! 09:09:06 we also want to use ironic integrating with Arista VXLAN-in-hardware, not there yet 09:09:49 hopefully have something to talk about on the Lustre front by Boston time 09:10:15 when you say "subnet mounts" do you actually mean "subtree mounts" ? 09:10:27 or is this another new lustre thing i don't know about? 09:10:53 doh yes, you're right 09:11:15 pretty quiet in here so far, i suppose this is non-US week though 09:12:14 I'm burbling away, don't let me get in the way of the agenda 09:13:09 daveholland, re. vxlan in hardware, are you talking about some sort of automated VNI per tenant with hardware VTEP support so your hosts just think they are on a native port ? 09:14:05 (always happy to burble) 09:14:31 there is really one main agenda item... 09:14:35 yes, we want hardware instances to get an untagged switch port, which the switch has on some automagically allocated internal VLAN, which is connected to the tenant network XVLAN mesh usnig the hardware VTEP 09:14:48 #topic Forum session submissions 09:15:41 we've got it working by manually the switch config (so we know bringing the XVLAN out to the hardware works) but it's the automation that lacks. AIUI. It's not my main focus a the moment 09:16:06 right, so there is a one-to-one VNI-to-VXLAN mapping for those VNIs so that they can be routed upstream somewhere? 09:16:56 similar to what we are doing on Mellanox+Cumulus gear actually 09:17:18 yes, the Arista gear talks to OpenStack so it's the same VNI that OVS uses 09:17:34 hello 09:17:46 only we haven't automated the VNI creation yet either 09:17:49 hi zioproto 09:17:59 ok, let's talk Forum 09:18:11 please open the etherpad 09:18:13 #link https://etherpad.openstack.org/p/BOS-UC-brainstorming-scientific-wg 09:19:34 aim here is to get names against items - these people will then take on submitting the session proposal and probably (unless unwilling or wanting assistance) (co-)chairing the sessions 09:20:07 I cant chair any session this time, because I will not be in Boston, sorry 09:21:48 zioproto, ah yes i remember. that's a shame! but understandable given ridiculous politics 09:24:50 zioproto, do i recall correctly that in Barcelona you put your hand up to lead any activities in the public dataset sharing space ? 09:25:53 sorry high prio phone call I have to leave the IRC meeting 09:26:12 np 09:27:19 is "Special hardware configurations" a new item in the list ? 09:27:51 i wonder if it could be merged with hypervisor tuning? different, but related... 09:29:50 looking at the etherpad. I can see new things in there we (Sanger) want but 1. we're not running bleeding-edge, still on Liberty, and 2. don't feel like enough of an expert to be proposing/chairing (sorry!) 09:31:21 daveholland, totally understand - i'm not claiming to be an expert either! trunk runs too fast and i can no longer keep-up with reading openstack-dev 09:32:25 though luckily i don't think expertise is necessarily a prerequisite, the reality is that a big chunk of deployments are still on N-2 N-3 releases 09:34:18 part of the ops and users bits in the summit/forum is just getting in a room with people who may know some of the answers - i think a session that answered questions common to many users by elucidating an existing upstream feature would still be considered valuable 09:36:16 ok i'm going to work through the etherpad now and fire off a few emails to folks who might be interested in pushing along some of these session topics - please mention me here to get my attention again 09:53:33 any other business before we time-out? 09:54:53 nothing from this end, sorry we won't see you in Boston 09:57:35 oh, you too :-( 09:57:50 will be very interesting to see what attendance is like 09:58:46 trump will make the data quite difficult in terms of evaluating the new Forum format 09:58:53 :( 09:59:47 ok thanks for coming daveholland (and staying the extra hour!) 10:00:04 #endmeeting