Monday, 2016-04-18

*** toddjohn_ has quit IRC00:47
*** kzaitsev_mb has joined #openstack-app-catalog00:50
*** kzaitsev_mb has quit IRC00:54
*** kzaitsev_mb has joined #openstack-app-catalog01:50
*** kzaitsev_mb has quit IRC01:55
*** kebray has joined #openstack-app-catalog02:37
*** kebray_ has quit IRC02:41
*** kzaitsev_mb has joined #openstack-app-catalog03:52
*** kzaitsev_mb has quit IRC03:57
*** kzaitsev_mb has joined #openstack-app-catalog04:53
*** kzaitsev_mb has quit IRC04:58
*** kzaitsev_mb has joined #openstack-app-catalog05:54
*** kzaitsev_mb has quit IRC05:58
*** kzaitsev_mb has joined #openstack-app-catalog06:54
*** kzaitsev_mb has quit IRC06:59
*** pcaruana has joined #openstack-app-catalog07:15
*** kzaitsev_mb has joined #openstack-app-catalog07:55
*** kzaitsev_mb has quit IRC08:00
*** kzaitsev_mb has joined #openstack-app-catalog08:01
*** kzaitsev_mb has quit IRC08:06
*** kzaitsev_mb has joined #openstack-app-catalog09:22
*** kzaitsev_mb has quit IRC10:24
*** kzaitsev_mb has joined #openstack-app-catalog11:01
*** kfox1111_ has joined #openstack-app-catalog11:04
*** kfox1111 has quit IRC11:06
*** toddjohn_ has joined #openstack-app-catalog12:20
*** toddjohn_ has quit IRC12:24
*** toddjohn_ has joined #openstack-app-catalog12:40
*** kebray has quit IRC14:36
*** kebray has joined #openstack-app-catalog14:40
openstackgerritStephen Gordon proposed openstack/app-catalog: Update to Fedora 23 Atomic 20160405 release
*** kebray_ has joined #openstack-app-catalog14:57
*** kebray has quit IRC15:00
*** spzala has joined #openstack-app-catalog15:18
* kfox1111_ is sad15:33
kfox1111_due to all the constraints of openstack templating, all we ever see is "single server to do..."15:33
*** reed has quit IRC15:34
*** reed has joined #openstack-app-catalog15:34
kfox1111_and the nova folks have turned up their noses at instance users, again.15:34
kfox1111_docaedo: I'm starting to think that the TC is the only org that has a hope of solving this.15:35
kfox1111_do you think we should bring the TC in on it? Otherwise, I think the app catalog will continue to struggle.15:36
docaedoI am not opposed to involving the TC in anything, I don't think having instance users would have any measurable impact on the app catalog15:39
kfox1111_why does everyone thing secret storage and retrieval is unimportant to applications?15:41
kfox1111_its usually less important for single server deployments, but is critical for multiserver deployments.15:41
kfox1111_its why we tend to only see "single server template for X..." getting contributed, instead of something that actually needs a cloud.15:41
docaedoI would venture it's because the openstack devs working on openstack servers are not building openstack applications15:41
docaedoSo people are definitely using heat templates to solve complicated multi-server problems15:43
docaedobut I don't think the templates they are building are useful for other people in a generic sense15:43
kfox1111_but not in a way that is contributable.15:43
docaedoinstance users, conditionals, something else would help15:43
kfox1111_which hurts the app catalog.15:43
kfox1111_all of those are important.15:43
kfox1111_and I'm loosing faith that openstack will address the issues before openstack becomes largely irrelevant. :/15:43
kfox1111_it has folks attention now. if it doesn't address the elephants in the room, someone else will. (and are)15:44
docaedoI think my view would be that the most important thing those impact will be the application developer ecosystem more so than the app catalog.  The ecosystem has to exist fist really15:44
kfox1111_agreed. and its unhealthy. :/15:44
kfox1111_which leads to an unhealthy app catalog. :/15:44
kfox1111_single server apps are easy to write. most folks can just do that themselves.15:45
kfox1111_its the multiserver ones that are harder, and need a real cloud dev to do. and benefit most from a cloud by adding scaling and fault tollerence.15:45
docaedodepends on the audience you're serve - there's a *much* larger audience for whom single server apps are mysterious and scary15:46
kfox1111_but those can't be contributed/easily consumed under the current model.15:46
kfox1111_but *pets*15:46
kfox1111_we tell users keep your *pets* off the cloud. and then we have an app catalog that favor's them. :/15:46
kfox1111_how can our users trust us when we start off with a fundimental assumption counter to the one made at the heart of what a cloud is about?15:48
docaedoon the other hand, does AWS have an equivalent?  A place where you can find multi-server "cloud apps"?15:48
kfox1111_if we wait for amazon, then we will always be following amazon, and they eventually will rule the world of clouds (they are pretty close already)15:49
kfox1111_they really don't care if people throw pets up because they still can kill them and get the mony for them existing for a while.15:50
kfox1111_it makes for a worse user experience, but they can claim, you've been warned, and they get more money by at least having lots of pets in the cloud for a while.15:50
kfox1111_openstack's not about that.15:51
kfox1111_or shouldn't.15:51
docaedowell the point I'm making isn't so much that we need to copy amazon - it's that it's a "market" that doesn't really exist, so even ignoring amazon, does anyone have a "marketplace" for cloud-scale apps?15:55
kfox1111_it can't exist until someone removes enough of the road blocks preventing it. once those are gone, the market will be created, and the first cloud there will have a huge leg up over the others.15:56
kfox1111_this has happened time and time again in capitalism.15:56
kfox1111_the old guard doesn't think a market is important. they don't do anything to let it be.15:57
kfox1111_someone new comes in and makes something easy that use to be hard.15:57
kfox1111_a new market flurishes.15:57
kfox1111_the old guard scramble to catch up, and some die.15:57
kfox1111_The cloud itself is one example of it.15:58
kfox1111_but then we so quickly forget that we have to innovate and push forward, or someone else will.15:58
docaedoCouldn't a cloud-scale app like this be packaged/managed via ansible?16:01
*** kebray_ has quit IRC16:01
kfox1111_its more possible to write a cloud scale app with ansible, but can't do stuff like autoscaling.16:01
kfox1111_I think ansible's grown out of the non cloud world, and adapted pretty well to living in the cloud world, but it still has some fundimental assumptions in it that make it a less then ideal ultimate solution.16:02
kfox1111_if you use it in a centrial place to control the nodes,16:02
kfox1111_you can rely on securtiy of your secrets, but then you can't do stuff like autoscaling, self healing, etc because a human has to be there.16:03
kfox1111_if you move ansible to the individual nodes, then you need instance users to provide security, and then autoscaling works.16:03
docaedoI agree that all of this stuff would be awesome to have for OpenStack :)16:04
docaedoSorry if I sound overly negative or pushing back too much on this16:04
kfox1111_sokay. It seems like only those who have tried to make autoscaling templates have tripped across these things. Then push hard for, why do we need such and such?16:05
kfox1111_the same exact thing was said about the cloud itself. why is the old way not good enough?16:06
docaedoJust think there are a bunch of fundamental things that need to come together to make it happen, I think the conversation/consideration is even bigger than instance users16:06
kfox1111_well, it mostly is good enough, except for ...16:06
kfox1111_but I'm not sure how to get there , when our attention span is usually 10 minutes, and tops out at 50 minutes at the summit.16:06
docaedohonestly I think Randy Bias was hitting on some of this like two years ago (digging for the post)16:06
kfox1111_it comes up periodically, then people push back, and folks give up, until someone new comes along and the cycle repeats.16:07
docaedoit's not the whole solution exactly, but points at the same symptoms and talks about what could help16:08
docaedoit's about that disconnect between what the devs are creating with all their openstack projects, and what the users/consumers of openstack need most16:09
kfox1111_yeah. that.16:09
kfox1111_we need a TC that will lead at times when the individual projects can't see the forest through their tree.16:10
kfox1111_which is why I was thinking we might have to take the issue to them.16:10
kfox1111_cross project only seems to cover cases where all the silo's agree there is a problem.16:10
kfox1111_nova doesn't think there is a problem,16:11
kfox1111_or at minimum doesn't believe it is their problem to care about.16:11
docaedoI think solving this is in the User Committee's charter more than anyones16:14
kfox1111_yeah, but where does the user committee get its power? do they have any other then to make suggestions?16:17
docaedoand big overlap with the interop efforts as well16:17
kfox1111_cuase suggesting stuff hasn't been working...16:17
docaedonobody has any more power than making suggestions, do they?16:17
kfox1111_to be fair, neither has putting up code. :/16:17
kfox1111_the tc does.16:17
kfox1111_all the silo's had to submit to decisions of the TC.16:17
kfox1111_if the TC says, you need functionality X for the better good of openstack, even though it makes your project a bit harder, then they are obligated to allow the functionality in.16:17
jrollare they, though? what's the consequence of not allowing it, even after the TC says they should?16:18
kfox1111_I believe they can kick out the ptl until they find one that will allow it.16:19
kfox1111_or kick the project out of the tent.16:19
jrollI can only imagine the storm that would happen if that happened16:19
jrollespecially booting nova out of openstack, heh16:19
kfox1111_they must submit to oversite by the TC as a condition to being in openstack.16:19
kfox1111_yeah, so the TC shouldn't abuse it.16:19
kfox1111_and the projects should think very carefully when the TC is willing to actually say they need X or Y. because they won't do it very often.16:20
kfox1111_heh... the [app] workgroup seems to have a very different definition of app then i do. :/16:21
*** pcaruana has quit IRC16:21
kfox1111_their examples are all "I want to write a program that uses the openstack api"16:21
kfox1111_not, "I want a cloud scaled application that runs on openstack that is easy to deploy as a unit (app)"16:22
docaedothink is the TC would only wield that kind of power in an extreme circumstance - if the user committee and TC can be convinced of a clear vision of how things could be implemented16:22
docaedoand yes, I believe your view of an openstack application is a little different from the consensus view16:23
kfox1111_to both.16:23
kfox1111_my defnition is important too. its just a few years ahead of where a lot of folks are thinking, I think.16:24
kfox1111_unfortunately, I few years will probably show us ubernetes in kubernetes plus multitenancy.16:24
kfox1111_and at that point, why bother deploying openstack? it will be easier to deploy kubernetes. :/16:24
kfox1111_for those with pets that are in the cloud, then openstack will probably become another cloud app (by my definition) deployed by kubernetes templates using kolla.16:25
kfox1111_and nich itself in for those who only care about iaas. which will be a diminishing number of users.16:25
docaedotoo many turtles there for me!16:25
kfox1111_k. well, in a few years, we'll see. :)16:26
kfox1111_containers really are changing the cloud world drastically.16:27
docaedoThat I completely agree with16:27
kfox1111_its very interesting to see the new cloud verses the old cloud, when the old cloud really isn't that old.16:27
docaedofor sure interesting times16:27
kfox1111_the old cloud still thinks its safe because its 'new'.16:27
kfox1111_But it feels like they are anoying those that need the new cloud use case, and are pushing those developers out. which means the new cloud stuff gets more developers. :/16:28
kfox1111_I still think openstack has a huge role to play, if they don't loose site of what makes the cloud awesome. but I'm afraid that's been lost more to visions of providing a vmware alternate.16:29
*** kfox1111_ is now known as kfox111116:32
docaedowell openstack isn't in danger of losing, all of the massive deployers who need to manage diverse infrastructure are choosing OpenStack.  But they're choosing it because they need an invisible IaaS so they can get resources via API16:33
docaedothey're not counting on it also providing a perfect application layer, as their users aren't even comfortable getting VMs from an API call yet let alone re-thinking their entire app architecture16:34
docaedo"massive developers" is not the right phrase, maybe I mean "massive organizations" (telcos, companies like walmart, etc)16:35
kfox1111yeah. the teloc's and walmart etc, will be just fine with openstack.16:44
kfox1111I think teloc's will have to continue with it, since their use case is really anticontainer friendy at the moment.16:45
kfox1111walmart on the other hand, could probably mostly drop openstack in a hot second and use something like kuberenetes.16:45
kfox1111it was built for scalable web apps.16:45
kfox1111openstack could see one if its beggest users mostly vanish in between 2 summits.16:46
kfox1111same thing happened with openstack in the first place. they want from like 10% usage of openstack, to 100% very quickly.16:46
docaedonah, kubernetes solves some problems but it's not a panacea, and it's not an IaaS16:47
kfox1111yeah, b ut they don't need iaas.16:47
kfox1111thats the point.16:47
kfox1111for example, we're running our physics cloud with one vm per compute host.16:47
kfox1111a couple hundred servers.16:47
kfox1111we then moved the compute task into a container on that vm, because we had issues with not having the exact same compute environment on all hosts.16:48
kfox1111(version skew)16:48
kfox1111likely, if kubernetes matures in the next 6 months, we will remove the vm layer, and just go back to physical hardware with the containers on top.16:48
kfox1111so our openstack cloud will shrink to about 9 servers, and the rest will be in a kubernetes managed cluster.16:48
kfox1111and of the remaining 9 servers,16:49
kfox1111about 40% of the vm's on them contain their servers entirely in containers, which could be moved out too....16:49
kfox1111the problme with IaaS is that it provided a great api for cloud app developers to treat things that use to not be automatable, as automatable.16:50
kfox1111but wasn't a perfect match. iaas refused to provide the api's they really cared about. for example, instance users.16:50
kfox1111but now that containers are a thing, and those systems are starting to provide what the users truely want, iaas will shrink.16:51
kfox1111for example:
kfox1111its a thing, and its a thing *today*.16:52
docaedoIt's possible but I still don't think the industry is going to move so quickly towards containers that it'll kill off big OpenStack deployments16:52
kfox1111barbican's a thing, but its a thing, maybe a year from now.16:52
kfox1111this is why I think openstack's niching itself into a corner. because one of my primary clouds is witnessing it first hand. openstack was 100% usage. it may quickly become 10% or less.16:53
kfox1111yeah, I think openstaack will stick around in a lot of places. but it won't be the crown jewel anymore. thats if openstack doesn't start making some important changes.16:54
kfox1111it still totally could. :/16:54
kfox1111but doesn't seem to want to. :/16:55
kfox1111I really really want to see openstack succeed in the cloud world. but I'm seeing more and more failed promices from it.16:55
kfox1111and nova's top of the list feature for mitaka is "Live migration improvements"16:56
kfox1111the top pet request of them all.16:56
docaedoWonder if this is remotely applicable to the solution? (haven't looked closely TBH but "users self managing roles without being admin" sounds promising)16:56
kfox1111yeah. last summit, there was a lot of discussion at the instance user talk about coming up with a new proxy service to wrap around all the other projects since they couldn't be bothered with adding securtiy to their own projects....16:57
kfox1111as an op, that scares me big time.16:58
kfox1111now you just add a whole nother layer of securty handling code that will contain bugs and can potentially be a security exploit waiting to happen. :/16:58
kfox1111we're now coming up with technical solutions to political problems! which is usually the death nell of projects.16:59
kfox1111they basically have to reimplement every projects api in their own api, with security extensions. the risk of them getting out of sync is very large. :/17:00
*** toddjohn_ has quit IRC17:11
*** toddjohn_ has joined #openstack-app-catalog17:12
*** toddjohn_ has quit IRC17:17
*** kzaitsev_mb has quit IRC17:28
*** kairat_ has joined #openstack-app-catalog17:34
*** kairat_ has left #openstack-app-catalog17:35
*** kairat_ has joined #openstack-app-catalog17:35
*** kairat_ has left #openstack-app-catalog17:35
*** kairat_ has joined #openstack-app-catalog17:36
*** kairat_ has left #openstack-app-catalog17:36
*** toddjohn_ has joined #openstack-app-catalog17:49
docaedokfox1111: Also will be sure to get some time in Austin with you and David Flanders, he's heading up community stuff for the foundation and is leading the charge from their side with respect to the app dev community17:49
*** pvaneck has joined #openstack-app-catalog17:56
*** kzaitsev_mb has joined #openstack-app-catalog18:50
*** kzaitsev_mb has quit IRC19:36
*** kzaitsev_mb has joined #openstack-app-catalog19:58
*** kzaitsev_mb has quit IRC20:10
kfox1111docaedo: awesome. thanks. :)20:10
*** kzaitsev_mb has joined #openstack-app-catalog20:15
*** toddjohn_ has quit IRC20:21
*** toddjohn_ has joined #openstack-app-catalog20:22
*** kzaitsev_mb has quit IRC20:22
*** toddjoh__ has joined #openstack-app-catalog20:24
docaedokfox1111 kzaitsev_ws - any concerns/feedback for ? Looks pretty good to me20:25
*** kzaitsev_mb has joined #openstack-app-catalog20:26
*** toddjohn_ has quit IRC20:27
*** toddjoh__ has quit IRC20:29
*** kzaitsev_mb has quit IRC20:35
kfox1111docaedo: I don't see anything wrong with it with a quick glance.20:35
kfox1111the zip/yaml thing is a little unsettling though.20:42
kfox1111are the flags different on the cli depending on the type?20:42
kfox1111I'd still like to see the usage issues addressed quickly before we put too many thingy's in there.20:42
docaedowhat's unsettling? just two different formats where csar is the container20:48
kfox1111we really need to add an cli thingy like:
kfox1111do we have what we need to do so?20:49
docaedoit depends on how the user wants to actually consume it - we don't have that for heat or murano stuff20:51
*** kzaitsev_mb has joined #openstack-app-catalog20:51
kfox1111we kind of have it for murano. heat, I agree we should have that too.20:53
kfox1111right now, we have a really terrible ui on the web site usability wise. :/20:54
kfox1111I'm hoping to fix it as part of the common code fix too, but need enough information to know what the cli should look like.20:56
kfox1111I know what it should look like for heat. I think I know what it is for murano. no clue for tosca20:56
docaedoTosca usage:
kfox1111I see no reference to zip there.20:59
docaedoI disagree that the UI is terrible from a usability perspective - but it depends on what you are expecting from it. Sounds like you want the UI to handle everything for the user, so they can just blindly click around and things will magically be installed for them?20:59
kfox1111is it expected that the user would unzip it and run parts of it?20:59
*** kzaitsev_mb has quit IRC20:59
kfox1111computers are good at mindless repeditive tasks. humans are not. here... try one for me. run this command and return me the results without pasting:21:00
kfox1111echo f14fa88548464736bbcf53bb2b7b02de21:01
kfox1111the ui there is horible.21:01
kfox1111just a simple tweak, and it is at least 10% more likely the user can do the task.21:01
kfox1111echo f14fa885-4846-4736-bbcf-53bb2b7b02de21:01
kfox1111a little bit of usability tweaking can really go a long way to making something better.21:02
*** kzaitsev_mb has joined #openstack-app-catalog21:02
kfox1111now add the ability to paste... it becomes hopefully almost 100% :)21:03
docaedoI don't disagree on that, I just disagree on blocking adding content that's usable by a lot of people.  I honestly think the entire heat section should be removed from the catalog, it's a complete waste of space, does nothing useful for the catalog :)  But I can't fix Heat's problems from the app catalog, I can only index heat assets21:04
kfox1111heh. I think the definition of alot there is a bit overblown... :/21:05
kfox1111I don't think heat's section is all that different then tosca really.21:05
kfox1111heat's at least works with a click in horizon. tosca, I'm not sure what to do with. its not archetected right.21:05
kfox1111it really seems like it should be plugged into heat, as an alternate format to hot.21:05
docaedofair, but I think the tosca examples in that review are more useful than the heat ones at least, and there's an active effort from the TOSCA folks to be engaged and use this. Remind me how many heat people have been interested in helping?21:06
kfox1111that I agree with.21:06
kfox1111I get the point. I guess we're optomizing for different things.21:06
kfox1111I'm trying to optomize for user experience.  your trying to maximise content.21:07
kfox1111both are valuable.21:07
kfox1111For the record, I'm not blocking either. blocking and not pushing are different things.21:07
docaedoI agree too - also I'm not opposed to adding more to the UI to make assets easier to use, but I think the two efforts are independent21:08
docaedoyou're completely right, my use of blocking was unfair hyperbole, I apologize :)21:08
kfox1111one makes more work for the other. When you have little resource,21:08
kfox1111the incentive to get your content in can be enough to improve usability.21:08
kfox1111if let in easy, then the incentive goes away.21:08
kfox1111no worries. I'm not mad. just pointing it out. :)21:08
docaedoI'm glad we're talking about this stuff and considering it, I'm very hopeful we'll get a bunch more people involved in the conversation from the user committee side, and app dev eco side, etc.21:08
kfox1111+1. :)21:09
kfox1111we didn't allow the hello world tocsa app in right away either. we pushed back a bit and got the user interface stuff in at the same time.21:09
kfox1111and it got done. :)21:09
kfox1111though the horizon stuff hasn't...21:10
kfox1111(There's a start of a patch for that waiting review though)21:10
kfox1111I had to do it.21:10
kfox1111It just shows the asset thoguh. no way to actually consume it.21:10
kfox1111so I'm not sure its worth showing in the ui or not.21:10
*** spzala has quit IRC21:14
kfox1111right now, it unchecks the checkbox for tosca, since they are not launchable, but still allows the user to check the box, then search through them.21:14
kfox1111figured that might be a good middleground.21:14
docaedosounds like a good approach to me21:14
kfox1111but if the user finds something they like, there's not an easy way to have them deploy it.21:15
*** spzala has joined #openstack-app-catalog21:36
*** spzala has quit IRC21:36
*** spzala has joined #openstack-app-catalog21:36
openstackgerritKirill Zaitsev proposed openstack/app-catalog: Add glare artifacts plugin for app-catalog asset types
openstackgerritKirill Zaitsev proposed openstack/app-catalog: Add script to upload assets to glare
openstackgerritKirill Zaitsev proposed openstack/app-catalog: Allow using glare for v1 API call and add v2 proxy
openstackgerritKirill Zaitsev proposed openstack/app-catalog: Implement UI for listing artifact drafts for glare assets
openstackgerritKirill Zaitsev proposed openstack/app-catalog: Add UI for adding artifact drafts to glare
kfox1111docaedo: see the magnum cross COE talk is at the same time as the app catalog meeting?22:06
docaedoyep, saw that :(22:17
docaedoyep, saw that :(22:17
kfox1111yeah.... I gave them the minimum api I think we might be interested in. hopefully at least that gets considered.22:17
kfox1111..... this is.... interesting...
kzaitsev_mbit might be not too late to request some swap. I believe I saw a letter from fuel guys about it. (I might be wrong and they were swapping their own things)22:18
docaedothat ec2 driver is interesting22:22
kfox1111no idea how that would work with neutron tenant networks. I'm guessing it doesn't. :/22:23
kfox1111if they could solve that though, I'd me much more in favor to bursting to aws then I am now.22:23
docaedoyeah it can't unless it also pipes in a VPN to the private side of your AWS network (which is technically not that difficult)22:24
docaedoI think that approach trades a lot of technical debt for the occasional convenience of having an extra set of VMs somewhere else22:25
*** toddjohn_ has joined #openstack-app-catalog22:26
kfox1111yeah. I like k2k federation to a real openstack cloud much better.22:26
kfox1111but aws is requested a lot by name in proposals around here. :/22:26
kfox1111this at least provides enough to claim aws support while still being 100% openstack api.22:27
docaedoUnless I missed something, that's almost a year old though22:28
kfox1111yeah. I was surprised by that too. but it was mentioned in the mitaka nova release notes.22:29
*** toddjohn_ has quit IRC22:30
*** toddjohn_ has joined #openstack-app-catalog22:31
kfox1111hmm.. no, I was mistaken. I still had the liberty release notes up too. I think it must have been that I was looking at.22:32
*** toddjohn_ has quit IRC22:34
openstackgerritMerged openstack/app-catalog: Update to Fedora 23 Atomic 20160405 release
kfox1111glance tasks api was suposed to replace the v1 api for upload from url?23:10
kfox1111cause mitaka claims tasks api is now admin only by default...23:11
*** toddjohn_ has joined #openstack-app-catalog23:13
docaedogoodness so weird and such a user-hostile change. That one I still don't understand the thinking behind23:14
kfox1111so they deprecated importing an image from url in v1, by adding it as a task in the task api via v2.23:21
kfox1111then they removed it in favor of an import api.23:21
kfox1111but I can't find reference to that in the api docs anywhere.23:22
kfox1111what the heck.23:22
kfox1111see deprication notes second bullet here:
kfox1111uh oh...23:35
kfox1111they made the angular launch instance workflow the default in Mitaka Horizon...23:35
kfox1111I don't think we ever tested that code path with the horizon plugin.23:35
kfox1111It may or may not work. :/23:35
kfox1111heh... "All Volume related panels in Horizon that previously used the term GB and gigabyte have been replaced with GiB and gibibyte."23:47
kfox1111and fallout from that lawsuit continues...23:47

Generated by 2.14.0 by Marius Gedminas - find it at!