12:58:54 <csatari> #startmeeting Review of Dublin edge notes
12:58:55 <openstack> Meeting started Fri Apr 20 12:58:54 2018 UTC and is due to finish in 60 minutes.  The chair is csatari. Information about MeetBot at http://wiki.debian.org/MeetBot.
12:58:56 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
12:58:58 <openstack> The meeting name has been set to 'review_of_dublin_edge_notes'
12:59:26 <csatari> #link https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG
13:00:47 <ChrisPriceAB> Good afternoon csatari.
13:00:52 <csatari> Hi
13:02:12 <csatari> I plan to go though the Wiki page topic by topic and record the comments in actions.
13:02:26 <ad_ri3n_> Hi
13:02:28 <jdandrea> Sounds good. (Will there be a roll call?)
13:02:44 <csatari> Oh, yes :)
13:02:56 <csatari> #topic Roll Call
13:03:15 <csatari> #info csatari - gergely.csatari@nokia.com
13:03:25 <ad_ri3n_> #info alebre - adrien.lebre@inria.fr
13:03:30 <pcarver> #info Paul Carver
13:03:36 <jdandrea> #info jdandrea - jdandrea@research.att.com
13:03:51 <esarault> #info Eric Sarault - eric.sarault@kontron.com
13:04:01 <ad_ri3n_> #info s/alebre/ad_ri3n_
13:04:18 <dpertin> #info dpertin dimitri.pertin@inria.fr
13:04:28 <ChrisPriceAB> #info Chris Price - christopher.price@est.tech
13:04:29 <ad_ri3n_> sorry … stupid habit
13:04:37 <csatari> No worries :)
13:04:42 <rcherrueau> #info rcherrueau - ronan-alexandre.cherrueau@inria.fr
13:05:22 <csatari> Okay, let's jump into the document.
13:05:34 <ad_ri3n_> so the idea was to go through the Dublin PTG wikipage.
13:05:37 <ad_ri3n_> :D
13:05:40 <ad_ri3n_> #link https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG
13:05:52 <csatari> Yes, chapter by chapter.
13:05:52 * jdandrea (change topic?)
13:06:03 <csatari> #topic Review Intro
13:06:11 <csatari> Anything to the intro part?
13:06:20 <ad_ri3n_> not from my side
13:06:40 <csatari> Okay, moving forward.
13:06:51 <csatari> #topic Definitions
13:07:06 <ad_ri3n_> Original Site : The site to where the operator is connected
13:07:10 <ad_ri3n_> should be probably reworded as:
13:07:20 <ad_ri3n_> The site where the operation is performed/executed initially
13:07:34 <csatari> Okay for me.
13:07:55 <ad_ri3n_> (BTW where are we taking notes? are we using an etherpad or something else?
13:07:56 <ad_ri3n_> )
13:08:01 <csatari> #action reword "The site to where the operator is connected" to "The site where the operation is performed/executed initially"
13:08:06 <ad_ri3n_> :-)
13:08:09 <ad_ri3n_> efficient ;)
13:08:15 <javier_> #info jrbalderrama - javier.rojas-balderrama@inria.fr
13:08:16 <csatari> I hope, that the meeting minutes will be enough,.
13:08:20 <csatari> Let's see :)
13:08:28 <ad_ri3n_> So based on that I think we should also reword the Remote site by
13:08:57 <csatari> Okay, any proposals?
13:08:59 <ad_ri3n_> Site(s) that are involved in an operation launched from the Orginal one.
13:09:05 <ad_ri3n_> but not sure it is clear enough
13:09:13 <ad_ri3n_>13:09:31 <csatari> maybe affected instead of involved?
13:09:43 <ad_ri3n_> better yes
13:10:20 <csatari> #action reword Remote Site to "Site(s) that are affected in an operation launched from the Orginal one."
13:10:27 <ad_ri3n_> The items Application Sustainability/Site Sustainability should be close each other
13:10:45 <fdag> #info Francis Dagenais - francis.dagenais@kontron.com
13:11:14 <csatari> The definitions are in alphabetical order now.
13:11:26 <ad_ri3n_> ok
13:11:32 <csatari> Should I break the alphabetical order?
13:11:37 <csatari> ok
13:11:41 <ad_ri3n_> don't know which one is the best one.
13:11:59 <ad_ri3n_> But since we do not have too many definitions, sorting by semantic is probably more relevant?
13:12:22 <ad_ri3n_> that's all from my side for 'Definitions'
13:12:24 <csatari> I would prefer to keep the alphabetical order to aviod long discussions about semanthics
13:12:28 <esarault> FOr those that's aren't well versed it might be easier to be alphabetical
13:12:33 <ad_ri3n_> ok
13:12:36 <csatari> and I hope we will have more definitions.
13:12:37 <csatari> Okay
13:12:46 <csatari> Anyonelese on the definitions ?
13:13:25 <csatari> #topic Review of 3 Edge use cases
13:13:36 <ChrisPriceAB> hmm maybe
13:13:42 <ChrisPriceAB> ^^
13:13:54 <csatari> ChrisPriceAB> to definitions?
13:14:29 <ChrisPriceAB> Is "original site" the way we want to represent the "operative site".  I assume an "original site" could be anywhere and refers to the site where an operation is being performed that may impact multiple sites.
13:14:35 <csatari> I always wanted to try to switch back to a previous topic :)
13:14:41 <ChrisPriceAB> lol
13:14:52 * ChrisPriceAB loves being fashionably late to a topic
13:15:12 <ChrisPriceAB> I'm just not 100% sure we are using the right term.  but I lack a better idea.
13:15:35 <ad_ri3n_> In my mind original site could be anywhere
13:15:42 <ad_ri3n_> if you have three sites A, B, C
13:15:43 <ChrisPriceAB> original is rather concrete for what I assume is a transactional definition.
13:16:02 <ad_ri3n_> user a connects to A to launch a request (whatever the request),
13:16:05 <csatari> Original - remote sounds better for me than Operative - Remote, but I let this be decided by native english speakers.
13:16:08 <ad_ri3n_> this is the original site for user a
13:16:34 <ad_ri3n_> user b connects to C, then C is the original site for the b request
13:16:42 <ad_ri3n_> then the request can be mono site or multi sites
13:16:53 <csatari> Yes, but the definition is per operation "The site where the operation is performed/executed initially"
13:16:56 <ChrisPriceAB> anyway, it's not urgent.  The context is an operation, and the way "original" presents does not help the reader understand the meaning.
13:17:01 <ad_ri3n_> (in the latter case, then the request, which is multi sites, includes remote sites)
13:17:11 <ad_ri3n_> per operation = per API request
13:17:19 <ad_ri3n_> exemple: openstack server create ….
13:17:36 <ChrisPriceAB> it will probably create a situation where the author understands, but maybe not the reader.
13:18:01 <ad_ri3n_> if the operation is start a VM on A with an image from B (sorry for this red thread), then the original site is A while site B is a remote one.
13:18:12 <ad_ri3n_> Maybe we can add a small schema to clarify that point?
13:18:15 * ChrisPriceAB is one of those people who doesn't read the definitions section before reading the document.  :)
13:18:32 <csatari> For now should I add a note to the page about this dilemma?
13:18:45 <esarault> Original makes me think of the "Primary" site, if the relation is relative to the sender's point of view, it needs ot be clarified
13:18:46 <ad_ri3n_> not sure it is a dilemma
13:18:56 <ad_ri3n_> if we draw a figure?
13:19:09 <esarault> A diagram would resolve that uncertainty for sure
13:19:16 <csatari> Got it.
13:19:18 <csatari> Okay
13:19:19 <csatari> #topic Definitions
13:19:44 <ad_ri3n_> @dpertin do we have any figure related to that point?
13:19:50 <ad_ri3n_> from our side?
13:20:04 <csatari> #action Create a drawing to explain Original and Remote sites
13:20:20 <dpertin> let me check
13:21:28 <csatari> dpertin: if you have anything please send it to me ;)
13:21:40 <csatari> Okay, anything else to the Definitions?
13:21:42 * ChrisPriceAB will try to stop making things difficult.  (but may fail to do so)
13:21:49 * ildikov is sneaking in and hiding in the corner at the back of the room :)
13:22:13 <csatari> Moving forward
13:22:13 <csatari> #topic Review of 3 Edge use cases
13:22:17 <dpertin> csatari: sure
13:22:20 <esarault> If you need some help on "beautifying" it let me know. Could probably get someone doing something in Adobe quickly
13:22:35 <ad_ri3n_> :-)
13:22:42 <ad_ri3n_> ok nothing csatari from my side
13:22:59 <ad_ri3n_> the use-cases are I think correctly explained in the White Paper (at least from now ;))
13:23:02 <csatari> dpertin, esarault Thanks. I deffinetly need someone with beutifying capabilities ;)
13:23:05 <ad_ri3n_> s/from/for now
13:23:19 <csatari> ok
13:23:40 <csatari> #topic Review of 4 Deployment Scenarios
13:23:56 <csatari> Only chapte 4  for now.
13:24:03 <csatari> (the sentrence)
13:24:12 <ad_ri3n_> ?
13:24:39 <csatari> If you have comments specific to 4.1 that should go to that topic.
13:25:03 <ad_ri3n_> Small Edge: I 'm wondering whether we can have information related to the storage capacity?
13:25:19 <csatari> #topic Review of 4.1 Small edge
13:25:45 <ad_ri3n_> i.e. a single hyperconverge server (i.e. ceph for instance will be also deployed to store VM/containers/baremetal images…) ?
13:25:47 <csatari> This is all the info what I could collect from all the etherpads.
13:26:04 <ad_ri3n_> Maybe it can make sense to open questions?
13:26:07 <esarault> Maybe one thing abour small edge, most SSDs are 240GB, not 225GB.
13:26:20 <ad_ri3n_> at least that a note on that?
13:26:22 <esarault> And for the Maximum, if the tagret is a Xeon-D type or ARM processor, probably 16 cores could eb the max tagret
13:26:53 <csatari> #action Change 225 GB to 24o GM
13:27:06 <fdag> Just a quick comment about Edge use cases (section 3). I'd like to stress that while yes, high amount of data vs low latency is a challenge, predictive, guaranteed maximal response times should be there somewhere
13:27:50 <fdag> This heavily implies RT schedulers support
13:27:56 <csatari> @all Is it OK if I add the 16 cores to the max specs of the Small edge?
13:28:28 <csatari> And the processor types.
13:28:59 <csatari> #action add to the maximum hardware specs that the tagret is a Xeon-D type or ARM processor, probably 16 cores
13:29:45 <csatari> fdag predictive, guaranteed maximal response time is a requirement, not an use case. Here we should describe why do we need predictive, guaranteed maximal response time.
13:30:49 <csatari> I'm happy to add more use cases if you have anything in mind.
13:31:18 <csatari> Anything else to Small Edge?
13:31:29 <esarault> I'd also liek to point out that it's unlikely people will refresh every OpenStack release their Small edge appliance. What we're seeing mostly from Tier 1/2 is them starting to be comfortable with a 2 years update cycle, opened to discuss 1 year
13:32:07 <esarault> updating every release is aggressive and time consuming for them and they typically have the mindset of "If it ain't broken, don't touch it"
13:32:44 <csatari> esarault even if the upgrade will be possible remotely without any affect to the running workload?
13:33:03 <esarault> Yeah the challenge is, you bring risk into the mix, a chance to mess with your SLA
13:33:11 <ad_ri3n_> csatari:  did you consider my remarks regarding the containers/vm/baremetals image repository (and more generally  what are the requirements in terms of cloud functionality)
13:33:30 <csatari> esarault: Understood.
13:33:41 <fdag> Are we expecting a FW release cycle of approx. 1month? This generally means downtime which may impact your on premise equipment.
13:34:04 <csatari> Should I change the update frequency to 1-2 years, then?
13:34:07 <esarault> An upgrade a year is managable I think. Sounds simple if you have a box, but if you have 10 000 of them scattered across the US, updating every 6 months becomes a burden
13:34:13 <ad_ri3n_> ^^ maybe we have general notes? It seems such a challenge will be valid for every edge deployments?
13:34:28 <parus> On Small Edge, the section on Remote access/connectivity reliability: It sounds like we assume allways on.... Why 100% uptime? I thought we assumed WAN link could fail often!
13:34:52 <parus> perhaps we should have a note that small edge availability is low.
13:35:01 <parus> Upgrades + Failures
13:35:12 <ad_ri3n_> parus:  it depends from which side the endusers/devlops will be.
13:35:20 <parus> ?
13:35:31 <fdag> csatari: understood for the requirements vs use case, but I see it more as a 'needs' in that case (as with 'needs high amount of data' and 'needs low latency handling')
13:35:33 <ad_ri3n_> maybe the connectivity between the enduser and the small edge site can be quite stable
13:35:54 <ad_ri3n_> while the connectivity between the edge site and the rest of the edge infrastructure can be intermittent
13:36:39 <ad_ri3n_> csatari:  seems there are a couple of comments/questions on that part. should we open an etherpad and copy/paste  the text in order to refine it?
13:36:50 <parus> Ok. but Failure or Management Upgrades may cause downtime without a local backup.
13:36:57 <ad_ri3n_> (i.e. on the depployment scenarios section).
13:37:47 <csatari> fdag: Okay, let's open an ehterpad for the Edge Use Cases text.
13:37:54 <esarault> Given the case is a small CPE box, I don't think the expectation is to maintain uptime during upgrade of software nor firmware. Target would be to minize the downtime but I don't think 100% is feasible given it's only one unit in the Min/Max specs
13:38:02 <csatari> #topic Review of 3 Edge use cases
13:38:30 <csatari> #info the text needs refinement.
13:38:34 <esarault> It would definitely be expected in the Medium edge however
13:38:39 <parus> Spelling: Autonomous
13:39:36 <csatari> #link https://etherpad.openstack.org/p/Dublin-edge-notes-wiki
13:40:23 <csatari> #action Anyone who would like to make modifications on 3 Edge use cases go to https://etherpad.openstack.org/p/Dublin-edge-notes-wiki and add your proposals
13:40:46 <csatari> Okay, back to the reliability of Small Edge-s
13:40:55 <ad_ri3n_> can we do the same for the deployment scenarios section?
13:41:04 <ad_ri3n_> (I copied/pasted the current text).
13:41:06 <csatari> yes, someone already dud :)
13:41:09 <csatari> did
13:41:47 <csatari> #topic Review of 4 Deployment Scenarios
13:42:15 <esarault> First timer with EtherPad here, you propose to just update the pad with what we propose directly? No need for comments or anything else?
13:42:32 <csatari> #action Proposals about 4 Deployment Scenarios should be added to https://etherpad.openstack.org/p/Dublin-edge-notes-wiki
13:42:38 <esarault> Don't want to break your usual process ^^
13:43:35 <csatari> esarault: No, just add your text. If you would like to add a note prefix it with "Note:" and I will not copy that to the wiki.
13:44:33 <csatari> Okay, back to the reliability of the Small Edge-s
13:45:01 <csatari> Should I add anything more or differnet, than the current text "No 100% uptime expected."?
13:45:59 <csatari> Okay, I see a "No 100% uptime expected and variable connectivity (e.g. connected car)" in https://etherpad.openstack.org/p/Dublin-edge-notes-wiki
13:47:07 <csatari> #topic Review of 4.2 Medium edge
13:47:09 <esarault> About the hardware specs concerning the image repo, expect the small edge, being a single unit to likely only have 1x M.2 SSD or 1.8" SSD drive
13:47:38 * esarault added quantity to the Etherpad
13:48:15 <ad_ri3n_> csatari:  I just read you comment on the pad. how do you want to process, actually I asked the question twice on the IRC, not sure you saw it (if yes sorry, I apologize but I just want to keep traces on that remark ;))
13:48:53 <parus> I think we should envision the compute node scenario.
13:49:00 <ad_ri3n_> we also discuss the two connectivy aspects. i.e. connectivity between end-users and the edge site and connectivity between the different edge sites?
13:49:26 <esarault> Regarding Medium Edge, I think 2U should be the minimum specs. There are units out there with Xeon Scalable with 4 independant servers within 2U and also some hyperconverged appliance with up to 18 servers in a 2U unit.
13:49:26 <ad_ri3n_> 10 min left should we try to put all the questions we may have somewhere ?
13:50:00 <csatari> ad_ri3n_ You mean the "i.e. a single hyperconverge server (i.e. ceph for instance will be also deployed to store VM/containers/baremetal images…) ?" issue?
13:50:16 <ad_ri3n_> yes
13:50:43 <parus> ad_ri3n_: who is the end-user here? the user of the VM? or an openstack tenant
13:51:22 <ad_ri3n_> parus:  the devops who requests the provisionning of a new VM
13:51:31 <ad_ri3n_> so both
13:51:44 <csatari> ad_ri3n_ We can start recording the open issue in info statements.
13:51:57 <csatari> s/issue/issues/
13:51:58 <ad_ri3n_> can be the end-user with his/her smartphone that requires to launch a new VM in the edge site
13:52:40 <ad_ri3n_> or can be a VM that is already running on the edge site and that wants to scale the service by provisionnng a new VM (just basic examples that come to my mind).
13:53:03 <csatari> esarault: I think 2U-s are too big for a small edge :)
13:53:03 <parus> The end-user may be on his smartphone watching you-tube.... but it might be goodgle who would launch the VM on the web-site.
13:53:25 <ad_ri3n_> csatari:  not sure I'm following you but ok, you are chairing the discussion. So I do what you propose ;)
13:53:40 <csatari> ad_ri3n_, parus: Maybe we should add a definition for the end user
13:53:45 <fdag> csatari: The comment from esarault applies to Medium Edge
13:53:57 <fdag> ;)
13:53:58 <csatari> fdag: okay.
13:54:12 <esarault> whatever fadg said :)
13:54:15 <esarault> *what
13:54:17 <csatari> :)
13:54:43 <csatari> #topic Definitions
13:54:54 <parus> How much longer can you guys go on this meeting?
13:55:15 <esarault> +30m here
13:55:21 <fdag> +30m
13:55:24 <csatari> #action add a defininition for the ones who are interaction with the edge cloud infrastructure and for the ones who a re using the services running on top of the edge cloud infrastructure.
13:55:27 <parus> good with me.
13:55:39 <csatari> +30 OK for me too.
13:56:05 <csatari> #topic Review of 4.2 Medium edge
13:56:30 <csatari> @All can we change the mimium specs of Medium edge from 4U to 2U?
13:57:15 <ad_ri3n_> I will leave in 3 min
13:57:17 <esarault> I'll say yes but then that's voting for myself :p
13:57:19 <ad_ri3n_> sorry I have another meeting from my side
13:57:30 <ad_ri3n_> what will be the difference
13:57:33 <ad_ri3n_> between 2U and 4U
13:57:37 <csatari> ad_ri3n_ okay
13:57:46 <esarault> 2U tagrets multi-node systems
13:58:00 <esarault> also ensures you're able to fit more compute when replacing an old 6U unit that dates from the WW1
13:58:21 <ad_ri3n_> csatari:  would it be possible to five a follow up to this meeting (actually I have several questions regarding the different acronyms such as MVS…. discussed lated in the wiki page).
13:58:23 <esarault> also allows to think that within "one box" they can be more than 1 server
13:59:44 <csatari> ad_ri3n_ Yes, I plan to have more sessions until we reach the end of the doucment :) Also if you have questions or comments feel free to send them to the edge DL or to this IRC channel.
14:00:05 <ad_ri3n_> ok thanks for all the great work you are doing ;)
14:00:13 <ad_ri3n_> it is really great to see more and more people contributing to this subject
14:00:14 <ad_ri3n_> thanks
14:00:15 <ad_ri3n_> CU
14:00:16 <ad_ri3n_> ++
14:00:20 <csatari> Okay, I do not have any prefernce on this to I'm happy to change it to 2RU
14:00:30 <csatari> ad_ri3n_: Thanks, my pleasure.
14:00:57 <csatari> #action Change the Minimum hardware specs of Meduim Edge to 2RU.
14:01:33 <csatari> There is a comment in the etherpad: Expected frequency of updates to hardware:  5-7 years
14:01:44 <csatari> Anyone against this?
14:01:56 <esarault> Yes, that's me. That's the typical lifeyccle of a server from a fianncial standpoint
14:02:13 <esarault> amortization in done in 3-5 years, system keep running for a few years afterward and slowly gets replaced
14:02:19 <csatari> esarault okay, thanks for the info.
14:02:22 <esarault> also ties to Intel's warranty on embeedded processors
14:02:28 <csatari> About this one: "Expected frequency of updates to firmware: Never unless required to fix blocker/critical bug(s)"
14:02:55 <csatari> I think we have more and more blocker/critical bugs in firmwares.
14:03:09 <csatari> never sounds a bit too optimistic.
14:03:11 <esarault> The thing is it'll create downtime
14:03:31 <csatari> Yes, that is clear.
14:03:33 <esarault> and from what we so, customers rarely update if ever there BIOS/BMC once in production
14:03:47 <esarault> *their
14:04:06 <fdag> csatari : You are right, but then qualification cycles of a specific FW is so long, most providers decide to live with the quirks instead
14:04:30 <csatari> Okay
14:04:52 <csatari> Anyone against this: Expected frequency of updates to control systems (e.g. OpenStack or Kubernetes controllers): 12 to 24 months
14:04:53 <csatari> ?
14:05:03 <fdag> Unless it's a big *cough* security flaw *cough*
14:05:13 <esarault> and this feedback comes from seeing Tier 1 customer and major public cloud providers
14:05:20 <esarault> Yeah, Specter let's say
14:05:53 <csatari> can you solve Spectre with a firmware update?
14:06:52 <esarault> @fadg by enforcing default disabling of options allowing Specter/Meltdown exploits, correct?
14:06:59 <csatari> Isn't is a screwdriver kind of update if not solved in every software piece (however some parts of the CPU can be also updated as a software)
14:08:00 <csatari> Okay, I will copy the text from the etherpad to the wiki based on this discussion.
14:08:04 <csatari> Moving on .
14:08:21 <fdag> Disabling some BIOS/ME options will 'workaround' the issue, yes
14:08:37 <csatari> #topic Review of 5 Features and requirements
14:09:40 <csatari> If nothing, then
14:09:55 <csatari> #topic Review of 5.1 Architectural paradigms
14:10:35 <dpertin> what do you mean by cloud metadata?
14:10:36 <esarault> nothing here either I guess
14:11:14 <csatari> #topic Review of 5.2 Features
14:11:37 <csatari> #action There are no levels mentioned anymore. Remove it from the sentence.
14:12:04 <csatari> #topic Review of 5.2.1 Base assumptions for the features
14:13:20 <csatari> #topic Review of Elementary operations on one site
14:13:36 <fdag> Typo: Network unreability -> Network unreliability
14:13:49 <csatari> #action Network unreability -> Network unreliability
14:13:56 <csatari> fdag Thanks
14:14:05 <dpertin> csatari: what do type: MVS and Non-MVS mean in the document?
14:14:26 <csatari> dpertin: MVS is minimum viable soluiton
14:14:37 <dpertin> ok thanks
14:15:19 <csatari> Anything else to
14:15:35 <csatari> #topic Review of Use of a remote site
14:16:14 <csatari> #action remove "Level 1"
14:16:33 <csatari> #topic Review of Network unreability
14:16:52 <esarault> This is Nono-MVS?
14:16:57 <esarault> *Non
14:17:08 <csatari> yes, according to my first guess.
14:17:19 <csatari> I'm open to discussions.
14:17:51 <csatari> Without this we can still have an edge infrastructure if we have good networks :)
14:17:54 <esarault> What's the impact if this is not there?
14:17:59 <esarault> True
14:18:17 <esarault> Just curious what a connection drop of 15 seconds impact's will be
14:18:26 <csatari> Network issues will result in failed operations and maybe inconsistent config data in the edge cloud infra.
14:18:43 <esarault> Some the user "resends" the command
14:18:57 <esarault> No notion of command buffering to retry x amount of time until accepted or declined?
14:19:26 <esarault> So basically "manual automation" until implemented :p
14:19:27 <csatari> This is what I mean on "Have a policy for operation retries"
14:20:12 <fdag> csatari : great!
14:20:22 <esarault> Yeah fair enough :0
14:20:24 <esarault> :)
14:20:32 <csatari> In the final solution I would keep trying for ever, but with some smartness to aviod killing the network in the minue as it returned from an outage.
14:21:16 <esarault> Yeah otherwise it might flag it as DDOS
14:21:33 <csatari> Yes
14:21:41 <dpertin> furthermore a site is supposed to provide L1 operations, even if it is not able to contact other sites
14:22:12 <csatari> Yes, and we miss this form this section
14:22:26 <csatari> #action a site is supposed to provide L1 operations, even if it is not able to contact other sites
14:22:35 <esarault> On that point, it might be good to clearly list the levels if they are to be refered to. Otherwise we'll loose people on terminologies here
14:23:05 <csatari> Hehe, I just wanted to ask what are the L1 operations :)
14:23:47 <csatari> At the moment I do not have a clear list of operation in mind.
14:24:18 <esarault> No worries, just wanted to point out some of us are closer to the hardware than the network ;)
14:24:32 <csatari> :)
14:24:37 <csatari> Okay
14:25:17 <csatari> #topic Review of Containers
14:26:05 <dpertin> From my viewpoint, L1 operations are the ones already provided in OpenStack
14:26:32 <esarault> Regarding containers, are we expecting to leverage Kuryr for this?
14:26:45 <csatari> I would limit those operations.
14:27:28 <csatari> Eg.: I would not let to overwrite the metadata locally what is received from a "parent" edge cloud instance.
14:28:01 <csatari> esarault It is not clear at the moment. There are several architecture options for this.
14:28:04 <esarault> I'm sure there's folks on the Medium edge that'll expect a symbiosis with K8S
14:28:21 <dpertin> csatari: what do you mean by metadata? images, flavors, etc?
14:28:55 <csatari> dpertin yes, things like that. See https://wiki.openstack.org/wiki/OpenStack_Edge_Discussions_Dublin_PTG#Metadata_distribution
14:29:17 <dpertin> csatari: thanks
14:29:23 <csatari> #topic What's Next
14:29:52 <csatari> #info We reached Containers. More similar meetings to come.
14:30:06 <fdag> @all I'll need to leave now, thank you very much for this session, great work!
14:30:07 <csatari> #action Csatari to organize the next meeting.
14:30:17 <parus> Thank you all
14:30:17 <csatari> fdag: Thank you.
14:30:31 <esarault> Thanks all, this was great. Looking forward to the next meeting!
14:30:38 <csatari> Anything to he AoB section?
14:30:50 <dpertin> Thanks
14:30:53 <csatari> Okay
14:30:56 <esarault> A0B?
14:31:01 <csatari> Thanks all.
14:31:07 <csatari> esarault: Any other business.
14:31:15 <csatari> #endmeeting