| *** kzaitsev_mb has quit IRC | 00:27 | |
| *** reed_ has joined #openstack-app-catalog | 04:16 | |
| *** reed_ has quit IRC | 04:18 | |
| *** kzaitsev_mb has joined #openstack-app-catalog | 09:43 | |
| *** kzaitsev_mb has quit IRC | 12:47 | |
| *** jolsen has joined #openstack-app-catalog | 13:03 | |
| *** openstackgerrit has quit IRC | 13:31 | |
| *** openstackgerrit has joined #openstack-app-catalog | 13:31 | |
| *** kzaitsev_mb has joined #openstack-app-catalog | 13:36 | |
| *** kebray has joined #openstack-app-catalog | 15:02 | |
| kfox1111 | any news on the release group? | 15:33 |
|---|---|---|
| kfox1111 | I got the rdo rpm to build and work. I've started submitting it to delorean so it will build from trunk. but it will not have a good version number until we can push tags. | 15:34 |
| docaedo | No responses from Herman, path of least resistance is wait for me to get my PTL crown (I think that's what I get right? Gold, encrusted with jewels? Ooh maybe a scepter too!) ... then when it's official, I can clean up all the groups associated with the app catalog. either tomorrow, or monday | 15:44 |
| kfox1111 | alittle confused. when is the vote? | 15:45 |
| docaedo | elections run from 18th to 24th, but I *think* if there's only one candidate then I can start making requests in advance | 15:47 |
| kfox1111 | ah. ok. | 15:48 |
| *** kebray has quit IRC | 16:01 | |
| kfox1111 | the mailing list said nominations are closed. so I think that means your basically official right now? :) | 16:17 |
| *** kebray has joined #openstack-app-catalog | 16:21 | |
| kzaitsev_mb | kfox1111: do you test ui against master horizon? I'm having a js error | 16:52 |
| kzaitsev_mb | Object is not a function (near '..._interface.hide();}}...') | 16:52 |
| kzaitsev_mb | global code534bfb91dd3a.js:42598 | 16:52 |
| kzaitsev_mb | docaedo: congrats on becoming PTL ;) | 16:52 |
| kzaitsev_mb | kfox1111: yep, nominations are closed indeed | 16:53 |
| docaedo | kzaitsev_mb: thanks! | 16:53 |
| kfox1111 | something very close to master. (master keeps changing... :) | 16:54 |
| kfox1111 | can you pastebin what your seeing? | 16:54 |
| kzaitsev_mb | kfox1111: http://paste.openstack.org/show/466566/ | 16:56 |
| kzaitsev_mb | Tried to make sure, that I use the latest versions of everything | 16:58 |
| kfox1111 | hmm... not seeing that code in the rdo checkout I have.. let me pull a fresh trunk... | 16:59 |
| kzaitsev_mb | Shall we go to meeting-3 ? =) | 16:59 |
| kfox1111 | could you set COMPRESS_ENABLED=False in local settings and try again? it may help narrow it down. | 16:59 |
| docaedo | meeting time in one minute :) | 16:59 |
| kfox1111 | ah. k. | 16:59 |
| kzaitsev_mb | kfox1111: seems that COMPRESS_ENABLED=False solves the issue =) guess we should lint the code sometime soon | 17:04 |
| kfox1111 | uh... thats strange... | 17:24 |
| kfox1111 | I'm seeing a different error on the newest trunk. | 17:25 |
| kfox1111 | ok... so if I turn off compression, it works for me too... | 17:28 |
| kfox1111 | so something's broken when compression is enabled... | 17:28 |
| kfox1111 | I'm guessing its something like a missing ';' ? ... | 17:33 |
| kzaitsev_mb | looks like something like that | 17:42 |
| kfox1111 | probably should run eslint against it... | 17:42 |
| kfox1111 | and watch it explode. :) | 17:42 |
| kfox1111 | know off hand how to run it on a specific file? | 17:44 |
| kzaitsev_mb | oh, I actually have 2 points, that i'd like to note. Should've done that on the meeting, probably | 17:44 |
| kzaitsev_mb | kfox1111: eslint some-file =) | 17:44 |
| kfox1111 | i did an npm install eslint, which worked, but eslint --help cant find the command. | 17:45 |
| kzaitsev_mb | npm install -g eslint | 17:45 |
| kfox1111 | ah. thx. | 17:45 |
| kzaitsev_mb | you should use the -g to install CLI arguments ) | 17:45 |
| kzaitsev_mb | was about to post that as you asked ) | 17:45 |
| kfox1111 | that fixed it. thanks. :) | 17:45 |
| docaedo | kzaitsev_mb: what are the two points? | 17:46 |
| kfox1111 | uh... it didn't explode... | 17:46 |
| kzaitsev_mb | so the 1st point is — murano has been making some changes to it's platform and it's apps, that make them backward-incompatible with kilo. | 17:47 |
| kzaitsev_mb | I think that just yesterday I stopped one of the developers from trying to update all the apps = | 17:47 |
| kfox1111 | thats unfortunate. :/ | 17:47 |
| kzaitsev_mb | =) | 17:47 |
| kzaitsev_mb | and thus breaking everything | 17:48 |
| kzaitsev_mb | for stable/kilo users ) | 17:48 |
| kzaitsev_mb | well. current apps explicitly say, that they work for kilo | 17:48 |
| kfox1111 | Is there a good reason to break compatibility, or was it an oversight? | 17:48 |
| docaedo | that's just silly :) at the very least, if they have versions that work post kilo, upload new versions and leave the existing content alone | 17:49 |
| kzaitsev_mb | docaedo: exactly | 17:50 |
| kzaitsev_mb | that was the 1st point I wanted to bring up | 17:50 |
| kfox1111 | or make the new artifacts contain the old ones in such a way that they work with both, would be the best option... | 17:50 |
| docaedo | kzaitsev_mb: thanks for stepping in and suggesting a little sanity | 17:50 |
| docaedo | kfox1111: that's the best approach, but I don't think murano does that (would be a good feature to suggest) | 17:50 |
| kfox1111 | is it too late to ask them to fix it? | 17:51 |
| kfox1111 | it almost seems like a bug... | 17:51 |
| kfox1111 | or are they updating the packages with features that only exist in the newest version? | 17:51 |
| kzaitsev_mb | I guess, that we would have to utilize versioning in app-catalog sometime soon, to accomodate L-based apps | 17:51 |
| kfox1111 | and there's no way to make them optional? | 17:51 |
| kzaitsev_mb | kfox1111: the latter for the most part, yes. | 17:52 |
| kfox1111 | I see... | 17:52 |
| kfox1111 | something I was thinking about implementing but hadn't gotten to it, was some how determining what service version was running, | 17:52 |
| kfox1111 | and then filtering out artifacts that didn't match the minimum for that service. | 17:53 |
| docaedo | kzaitsev_mb: I think new assets can be uploaded before versions work by just using new names (like with liberty in the name) | 17:53 |
| kzaitsev_mb | things would break, because new apps, try to utilize features only available in liberty | 17:53 |
| docaedo | like "Apache Tomcat (liberty)" and note which release it works with in the asset | 17:54 |
| kfox1111 | we can look at the version of horizon, and guess the default version of the rest of the services, but it could be wrong. | 17:54 |
| kzaitsev_mb | docaedo: I guess for murano they might have versions concatenated to their names. | 17:54 |
| kzaitsev_mb | in case we end up versioning with something like semver | 17:54 |
| kfox1111 | this is kind of a facet of multiple versioning. | 17:54 |
| kzaitsev_mb | docaedo: that can work too | 17:55 |
| kzaitsev_mb | nice idea | 17:55 |
| kfox1111 | we may have multiple versions of a package, and for each package version, there may be multiple versions by which engine needs to run it. :/ | 17:55 |
| docaedo | kzaitsev_mb: for now it's a hack, but at least it doesn't break things for other people | 17:55 |
| kzaitsev_mb | so it is not a reall issue right now, but rather I just wanted to bring it up. | 17:55 |
| kfox1111 | there isn't a restriction curently that disallows multiple murano assets from implementing the same package_name... | 17:56 |
| kfox1111 | though.... | 17:56 |
| kfox1111 | I think murano's engine would break on that one. :/ | 17:56 |
| docaedo | and yeah, eventually we'll have to build in the concept of multiple versions of the same asset, and provide the ability to ask for an older version | 17:56 |
| kfox1111 | yeah, this is a murano issue.... | 17:56 |
| kfox1111 | since they pull stuff themselves from the catalog, their layout will have to adapt to having same version, different engines. :/ | 17:57 |
| kzaitsev_mb | kfox1111: there is also an issue, that murano assets are stored and fetched from the http server on the app-catalog side. | 17:57 |
| kfox1111 | the murano engine expects a package_name to be unique in the repo... | 17:57 |
| kfox1111 | exactly. | 17:57 |
| kzaitsev_mb | I believe this would be an issue for murano during M cycle | 17:58 |
| kfox1111 | so the murano engine will have to be changed to fix this. | 17:58 |
| kfox1111 | they need to decouple package_name from package_name in the app catalog... | 17:58 |
| kfox1111 | oh. liberty is not affected? | 17:58 |
| kzaitsev_mb | and I hope to add app-catalog API integration to muranoclient during M | 17:58 |
| kfox1111 | ok. then we need to make sure we roll this up into the versioning discussion. | 17:59 |
| kfox1111 | and think hard about it... | 18:00 |
| kzaitsev_mb | for L though =/ | 18:00 |
| kzaitsev_mb | well old apps should work fine on L | 18:00 |
| kfox1111 | cause murano package_name has to be unique within an instance of murano, but can't be in the catalog... | 18:00 |
| kzaitsev_mb | and if we simply add L-named apps — they should also work fine | 18:00 |
| kfox1111 | k. so they have to use new package_name's for liberty stuff, until we can get versioning streightened out with both sides. | 18:01 |
| kzaitsev_mb | kfox1111: not anymore (I mean inside murano) | 18:01 |
| kzaitsev_mb | I believe client is not yet aware of that though ) | 18:01 |
| kzaitsev_mb | one of the changes murano mad during L was versioning — allowing multiple versions of 1 package inside of it. | 18:01 |
| kfox1111 | yeah, I meant as of liberty, there's no way to tell murano to pull a package and its dependencies from the catalog, | 18:01 |
| kfox1111 | and have a kilo and a liberty package there with the same package_name. | 18:02 |
| kzaitsev_mb | yeah, there's probably no helping that =) | 18:02 |
| kfox1111 | so until that's fixed, the only option is to have different package_names for Liberty stuff. | 18:02 |
| kfox1111 | well... is this a bug and can they still fix bugs? | 18:03 |
| kfox1111 | do we still have time to slip in a fix that would work well enough? | 18:04 |
| kzaitsev_mb | a bug for the apps not updated in the app-catalog? | 18:04 |
| kfox1111 | no, a bug in murano engine not supporting versioning with the app-catalog. | 18:04 |
| kfox1111 | ie, if a liberty engine wants to import a package from the catalog, it should prefer a liberty tagged asset. | 18:04 |
| kfox1111 | wait a second... | 18:05 |
| kfox1111 | idea.. | 18:06 |
| kfox1111 | ok... so the murano package name comes from the catalog, and its handed to the murano engine for importing from the catalog... | 18:06 |
| kfox1111 | but its not apps.openstack.org, its something else... | 18:07 |
| kzaitsev_mb | (actually by murano client, but that doesn't change much =) ) | 18:07 |
| kfox1111 | a http site with a bunch of zip files in it, named by package_name... | 18:07 |
| kzaitsev_mb | yes, that's how things happen now. | 18:07 |
| kfox1111 | why dont we fork that site? | 18:07 |
| kfox1111 | and change the default repo url in liberty murano? | 18:08 |
| kfox1111 | we point it to the "Liberty" murano repo. | 18:08 |
| kzaitsev_mb | huh, nice idea. | 18:08 |
| kzaitsev_mb | something like http://liberty.storage.openstack.org | 18:08 |
| kfox1111 | yeah. | 18:08 |
| kfox1111 | or maybe make it a script, so that it can do things more dynamic? | 18:09 |
| docaedo | I would not say I'm firmly against that idea, but I'm definitely not in favor of it | 18:09 |
| kfox1111 | like storage.openstack.org/liberty | 18:09 |
| kzaitsev_mb | that would work too | 18:09 |
| docaedo | I like the second idea at least a little more (i.e. subdirectory in the files for per-release issues that murano has) | 18:09 |
| kfox1111 | the script can then based on the path, pick the most appropriate version of a package for the engine. | 18:09 |
| kfox1111 | it can be smart and union too. | 18:10 |
| kfox1111 | if there isn't a liberty tagged murano package, it can fall back to a kilo one. | 18:10 |
| kzaitsev_mb | kfox1111: I think, that if we put http://storage.openstack.org/liberty as the url, there would be no need to do any more work. | 18:11 |
| kfox1111 | Ideally, we should be able to do an http redirect to a file, so it gets cached correctly... | 18:11 |
| kfox1111 | kzaitsev_mb: true, except we'd have to copy every murano package to both sides. | 18:11 |
| kzaitsev_mb | kfox1111: cause I suspect the code that chooses the most appropriate branch would get thrown away at some point anyway =) | 18:12 |
| kfox1111 | yeah, but we probably want to support murano kilo/liberty for a few years. | 18:12 |
| kfox1111 | so we might have 3 versions for a while. | 18:12 |
| kfox1111 | kilo, liberty, and mitaka+ | 18:12 |
| kfox1111 | once we solve versioning in mitaka. | 18:13 |
| kfox1111 | ok... can you verify if the repo loading works with http redirects? | 18:13 |
| kzaitsev_mb | That's some nice idea for the current situation. | 18:13 |
| kzaitsev_mb | kfox1111: if it doesn't — it's a bug | 18:13 |
| kfox1111 | ok. so... | 18:13 |
| kzaitsev_mb | although that might mean, that current kilo installations might fail. | 18:14 |
| kfox1111 | how about, we make the repo endpoint apps.openstack.org/api/v1/murano_repo/liberty/ | 18:14 |
| kzaitsev_mb | I'll have to verify it anyway ) | 18:14 |
| kfox1111 | we leave storage.apps.openstack.org alone. it will work for kilo. | 18:14 |
| kfox1111 | at apps.openstack.org/api/v1/murano_repo/liberty/, we have a script. | 18:14 |
| kfox1111 | it looks in the json for a liberty tag'ed asset. if there is one, | 18:14 |
| docaedo | afaik murano never touches apps.openstack.org for anything, it expects to find its bits at storage.apps.openstack.org | 18:15 |
| kzaitsev_mb | oh, you want to make it app-catalog | 18:15 |
| kfox1111 | it redirects to the filename under storage.apps.openstack.org/libert/%packagename.zip | 18:15 |
| kfox1111 | otherwise, it redirects to storage.apps.openstack.org/%packagename.zip | 18:15 |
| kzaitsev_mb | docaedo: I think kfox1111 suggests we set MURANO_REPO_URL to "http://apps.openstack.org/api/v1/murano_repo/liberty/" | 18:16 |
| kzaitsev_mb | so murano would still not touch anuthing in app-catalog | 18:16 |
| kfox1111 | then, that. :) | 18:16 |
| kfox1111 | then the app catalog code can point murano at the correct version of a package. | 18:16 |
| docaedo | kzaitsev_mb: ah got it - I would think better to set MURANO_REPO_URL to storage.openstack.org/liberty | 18:16 |
| kfox1111 | and murano's none the wiser. :) | 18:16 |
| kfox1111 | problem with storage is its static. we can' | 18:17 |
| kfox1111 | t have a script there. | 18:17 |
| kzaitsev_mb | I'm a bit suspicious to adding this stuff to api too ) | 18:17 |
| docaedo | We shouldn't have to write a script on our side, as this really is a murano issue (and IMO murano folk should coordinate with us even just a tiny bit if they want murano assets to keep working in the catalog) | 18:17 |
| kfox1111 | I agree we shuldn't have to fix it on our end, but its not something that was forseen until it was too late. :/ | 18:17 |
| kfox1111 | its a pretty small fix on our end to work around. | 18:18 |
| kfox1111 | I think for now, the very quickest thing we can do, is make http://apps.openstack.org/api/v1/murano_repo/liberty/ redirect to storage.apps.openstack.org/ and get murano to change the default to the former. | 18:19 |
| kzaitsev_mb | kfox1111: we can put the script to the storage.openstack.org/liberty too, just a little more configuration there | 18:19 |
| docaedo | kfox1111: sure, if that approach does work and does not require anyone with murano to do anything, I'm all for it. I definitely will not stand in the way :) just saying, there's a limit to how much the app-catalog should do going forward (i.e. fine for a last minute thing like this, but if there's some new release that makes unstated assumptions on how the app-catalog will work, it's not going | 18:19 |
| kfox1111 | we can then optomize it later. | 18:19 |
| docaedo | to always be our problem) | 18:19 |
| kfox1111 | kzaitsev_mb: I thought storage was a swift endpoint? | 18:19 |
| docaedo | yeah, storage is a rackspace cloud files endpoint (so basically swift, except not exactly swift AFAIK) | 18:20 |
| kzaitsev_mb | kfox1111: hah, I believe it's currently just a dumb http thing, isn't it? | 18:20 |
| kfox1111 | docaedo: agreed. Its mostly caused by the lack of us having versioning. We should fix that for sure in Mitaka. | 18:20 |
| kzaitsev_mb | oh | 18:20 |
| kfox1111 | so no scripts in storage. | 18:20 |
| kfox1111 | ok. so, kzaitsev_mb, can you put up a patch to change the default url? | 18:21 |
| docaedo | kfox1111: yes, we can fix versions, but I'm saying murano may very well STILL not handle versions however we build them in the app catalog, unless murano wants to track this stuff. I will keep folk informed at any rate (and that reminds me, I should join the murano channel!) | 18:21 |
| kzaitsev_mb | =) | 18:21 |
| kfox1111 | agreed. | 18:21 |
| kfox1111 | we need a discussion between murano and app-catalog going forward during mitaka to make sure versions are implemented properly on both sides. | 18:22 |
| kzaitsev_mb | kfox1111: yep, patch would be pretty straightforward | 18:22 |
| kzaitsev_mb | do you want it, just to have a WIP version? | 18:22 |
| kfox1111 | yeah. I'll start the infra one. once the infra merges, then we should be able to murge the other. | 18:23 |
| kzaitsev_mb | ok. will do that bit later today. have to go now. Thanks for those insights, was really helpfull | 18:24 |
| kfox1111 | np. thanks for bringing up that problem. it would have bit lots of folks. :/ | 18:24 |
| docaedo | kzaitsev_mb: yes thanks | 18:25 |
| docaedo | When there's a WIP version, let me know, I (or one of you?) can send a note to ML highlighting the issue, and making sure there's some visibility there regarding how we can/should solve this down the road | 18:26 |
| kfox1111 | +1 | 18:26 |
| *** kzaitsev_mb has quit IRC | 18:29 | |
| kfox1111 | docaedo: https://review.openstack.org/224869 | 18:37 |
| kfox1111 | it currently makes the murano liberty engine function exactly like the kilo one. | 18:37 |
| kfox1111 | that will allow them to change the default to the new endpoint very quickly. | 18:38 |
| kfox1111 | we can fix it to be smarter when we get the first liberty murano package that conflicts with a kilo one in. | 18:38 |
| kfox1111 | that could be our first flask test. :) | 18:41 |
| docaedo | ok .. now, I suppose we need Murano folks to update all the packages and share the liberty versions right? | 18:53 |
| kfox1111 | add entries for new packages not already there, and give us liberty versions of the ones that are liberty only. | 18:55 |
| docaedo | wait, this would redirect http://apps.openstack.org/api/v1/murano_repo/liberty/someasset.zip to http://storage.apps.openstack.org/someasset.zip right? | 18:56 |
| kfox1111 | right. so as it stands, it works exactly as it does today. | 18:56 |
| kfox1111 | but, they can update the default url to http://apps.openstack.org/api/v1/murano_repo/liberty/ and it will function. | 18:57 |
| kfox1111 | then, we either do 1 of two things: | 18:57 |
| docaedo | ah well in that case wouldn't we want to redirect to storage.apps.openstack.org/liberty ? | 18:57 |
| kfox1111 | 1. copy everything in http://storage.apps.openstack.org/ to http://storage.apps.openstack.org/liberty, and update the redirect, | 18:57 |
| kfox1111 | or we put flask at http://apps.openstack.org/api/v1/murano_repo/liberty | 18:58 |
| docaedo | hmm, ok I think before we merge this or do anything then, I'll share the link with a note to the mailing list right now, and give them a chance to respone | 18:58 |
| kfox1111 | and have it intelengently issue redirects to either storage.apps.openstack.org/ or storage.apps.openstack.org/liberty based on if there is a kilo/liberty conflict. | 18:58 |
| docaedo | respond | 18:58 |
| docaedo | wdyt? | 18:59 |
| kfox1111 | what I would like to do is get something in place as soon as possible to fix the bug, before they release a "final version" that has a bad url it in. :/ | 18:59 |
| kfox1111 | we can fix it on our server at our leasure. | 18:59 |
| kfox1111 | we can't fix released client packages easily. :/ | 18:59 |
| docaedo | because at this point we're making some assumptions, maybe they want to handle it differently? I'm with you on getting it fixed, and for us, easy to fix whenever we want (ha yes, you just typed that) | 19:00 |
| *** rsFF has joined #openstack-app-catalog | 19:00 | |
| kfox1111 | Wouldn't hurt to merge the infra thing. if we have to update it again, thats ok. and it takes time to get the merge through ,so the soner the better. | 19:01 |
| kfox1111 | I expect the review on the murano side will spark the discussion. | 19:01 |
| kfox1111 | we can raise it on the mailing list too. | 19:01 |
| kfox1111 | Since it takes days simetimes to get stuff through infra, I figured we'd start that asap. | 19:01 |
| docaedo | I guess I'm confused then, because things work as-is, and that patch gives murano folk an option if they want to use it, but nobody from there has asked for it | 19:02 |
| kfox1111 | the issue is, | 19:02 |
| kfox1111 | if they release as it is today, without the new url, | 19:02 |
| kfox1111 | then liberty murano is stuck at kilo forever. | 19:03 |
| kfox1111 | se can't fix it. | 19:03 |
| kfox1111 | we | 19:03 |
| docaedo | ah ok so the proposition is: | 19:03 |
| docaedo | 1) merge this patch | 19:03 |
| kfox1111 | if we make it work just enough so that they can release with the differeent url, then we can fix it without having to update all the clients in the field (impossible) | 19:03 |
| docaedo | 2) tell murano if they have any hope of keeping things compatible, use the new repo URL and then we can make adjustments later as needed | 19:04 |
| kfox1111 | correct. | 19:04 |
| docaedo | got it, ok we are on the same page now.. whee! | 19:04 |
| kfox1111 | :) | 19:04 |
| docaedo | as long as I'm not distracted by something shiny in the next few minutes, I can write that up and send a note to the ML. Will also +1 that PR | 19:06 |
| kfox1111 | k. thanks. :) | 19:06 |
| docaedo | np, thanks for helping explain it to me, I clearly was not getting this at first | 19:07 |
| kfox1111 | np. :) | 19:08 |
| kfox1111 | kzaitsev_ws: I'm seeing the error with the compressed version whether or not I'm using the app catalog... I think something may be broken in trunk? | 19:33 |
| *** kebray has quit IRC | 19:48 | |
| docaedo | message sent (http://lists.openstack.org/pipermail/openstack-dev/2015-September/074842.html) and also posted a heads-up on the murano IRC channel | 20:05 |
| *** kebray has joined #openstack-app-catalog | 20:13 | |
| *** kebray has quit IRC | 21:00 | |
| *** kebray has joined #openstack-app-catalog | 21:00 | |
| *** kzaitsev_mb has joined #openstack-app-catalog | 21:39 | |
| kfox1111 | docaedo: cool. thanks. | 21:39 |
| *** kebray has quit IRC | 22:04 | |
| *** kebray has joined #openstack-app-catalog | 22:13 | |
| *** kzaitsev_mb has quit IRC | 22:19 | |
| *** kzaitsev_mb has joined #openstack-app-catalog | 22:21 | |
| *** kebray has quit IRC | 22:52 | |
| kfox1111 | kzaitsev_mb: try: | 22:56 |
| kfox1111 | git diff 4c6d88d38f8e9f9f5da0bf38c9b7564172d7b960..d6ce0cf9f0fb809b241ab68118cb47f722e6501b | patch -R -p1 | 22:57 |
| kfox1111 | In horizon trunk. | 22:57 |
| kfox1111 | I think thats the commit that is breaking compression. | 22:57 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!