Tuesday, 2015-09-22

*** achanda has quit IRC00:03
*** flwang has quit IRC00:03
*** achanda_ has quit IRC00:04
*** flwang has joined #openstack-glance00:04
*** kebray has quit IRC00:16
*** keekz has quit IRC00:24
*** keekz has joined #openstack-glance00:25
openstackgerritMerged openstack/glance_store: Handle swift store's optional dependency  https://review.openstack.org/20974400:26
*** stevemar has joined #openstack-glance00:27
*** dims_ has quit IRC00:42
*** nikhil has quit IRC00:45
*** harshs has quit IRC00:53
*** harshs has joined #openstack-glance00:53
*** sdake_ has joined #openstack-glance01:04
*** sdake has quit IRC01:06
*** sdake_ has quit IRC01:07
*** sdake has joined #openstack-glance01:07
*** smatzek has joined #openstack-glance01:18
*** dims has joined #openstack-glance01:26
*** mingdang1 has joined #openstack-glance01:43
*** stevemar has quit IRC01:46
*** haomaiwa_ has joined #openstack-glance01:51
*** haomaiwa_ has quit IRC02:01
*** 6A4AAH74W has joined #openstack-glance02:01
*** smatzek has quit IRC02:01
*** wokuma has left #openstack-glance02:15
*** mtreinish has quit IRC02:28
openstackgerritMerged openstack/glance: Add unit tests for signature_utils class  https://review.openstack.org/21472602:33
*** achanda has joined #openstack-glance02:33
*** dims has quit IRC02:33
*** sdake has quit IRC02:39
*** david-ly_ has joined #openstack-glance02:39
*** gberginc_ has quit IRC02:40
*** david-lyle has quit IRC02:40
*** harshs has quit IRC02:40
*** dhellmann has quit IRC02:40
*** dhellmann has joined #openstack-glance02:41
*** baojg has joined #openstack-glance02:43
*** tobe has joined #openstack-glance02:48
*** tobe has quit IRC02:48
*** mingdang_ has joined #openstack-glance02:50
*** mingdang1 has quit IRC02:52
*** mingdang_ has quit IRC02:52
*** mingdang_ has joined #openstack-glance02:52
*** stevemar has joined #openstack-glance02:56
*** mingdang1 has joined #openstack-glance02:56
*** mingdang_ has quit IRC02:57
*** mtreinish has joined #openstack-glance02:59
openstackgerritKent Wang proposed openstack/glance: Allows OVA/OVF package upload to Glance  https://review.openstack.org/21481002:59
*** 6A4AAH74W has quit IRC03:01
*** haomaiwa_ has joined #openstack-glance03:01
*** itisha has joined #openstack-glance03:04
*** gberginc has joined #openstack-glance03:07
*** kebray has joined #openstack-glance03:10
*** kebray has joined #openstack-glance03:11
*** david-ly_ is now known as david-lyle03:17
*** baojg has quit IRC03:22
*** nikhil has joined #openstack-glance03:22
*** ayoung has quit IRC03:30
openstackgerritOpenStack Proposal Bot proposed openstack/glance: Updated from global requirements  https://review.openstack.org/22461003:32
openstackgerritOpenStack Proposal Bot proposed openstack/glance_store: Updated from global requirements  https://review.openstack.org/22461103:32
*** dims has joined #openstack-glance03:34
openstackgerritOpenStack Proposal Bot proposed openstack/python-glanceclient: Updated from global requirements  https://review.openstack.org/22466803:36
openstackgerritMerged openstack/glance: Fixed non-owner write-access to artifacts  https://review.openstack.org/21837903:58
*** haomaiwa_ has quit IRC04:01
*** haomaiwa_ has joined #openstack-glance04:01
*** harshs has joined #openstack-glance04:08
*** dims has quit IRC04:09
*** stevemar has quit IRC04:09
*** stevemar has joined #openstack-glance04:10
openstackgerritMerged openstack/glance: Fixed an HTTP 500 on artifact blob upload  https://review.openstack.org/21618004:15
openstackgerritMerged openstack/glance: Prevent extraneous log messages and stdout prints  https://review.openstack.org/22478704:16
*** mtreinish has quit IRC04:18
*** gberginc has quit IRC04:22
*** mtreinish has joined #openstack-glance04:23
*** lakshmiS has joined #openstack-glance04:34
*** harshs has quit IRC04:36
*** haomaiwa_ has quit IRC05:01
*** haomaiwang has joined #openstack-glance05:01
*** changbl has quit IRC05:11
*** changbl has joined #openstack-glance05:15
*** abhishekk has joined #openstack-glance05:19
*** dsariel_ has joined #openstack-glance05:20
*** mingdang1 has quit IRC05:36
*** kebray has quit IRC05:37
*** stevemar has quit IRC05:53
*** gberginc has joined #openstack-glance05:55
*** haomaiwang has quit IRC06:01
*** haomaiwang has joined #openstack-glance06:01
*** gberginc has quit IRC06:12
*** groen692 has joined #openstack-glance06:16
*** sabari has left #openstack-glance06:29
*** itisha has quit IRC06:33
*** gberginc has joined #openstack-glance06:34
*** itisha has joined #openstack-glance06:36
*** mingdang1 has joined #openstack-glance06:52
*** sabari has joined #openstack-glance06:53
*** david-lyle has quit IRC06:58
*** haomaiwang has quit IRC07:01
*** haomaiwang has joined #openstack-glance07:01
openstackgerritSabari proposed openstack/glance: Add config option for importing subflows  https://review.openstack.org/19489807:05
*** david-lyle has joined #openstack-glance07:14
*** jamielennox is now known as jamielennox|away07:15
*** dsariel_ has quit IRC07:16
*** sgotliv has joined #openstack-glance07:16
*** achanda has quit IRC07:17
*** takedakn has joined #openstack-glance07:28
openstackgerritMerged openstack/glance: Use dictionary literal for dictionary creation  https://review.openstack.org/21030307:32
openstackgerritMerged openstack/glance: reuse the deleted image-member before create a new image-member  https://review.openstack.org/19089507:32
*** mingdang1 has quit IRC07:35
*** DuncanT has quit IRC07:41
*** itisha has quit IRC07:41
*** serverascode has quit IRC07:41
*** marcusvrn_ has quit IRC07:41
*** boris-42 has quit IRC07:41
*** zhiyan has quit IRC07:41
*** ameade has quit IRC07:41
*** briancurtin has quit IRC07:41
*** briancurtin has joined #openstack-glance07:43
openstackgerritAlexander Tivelkov proposed openstack/glance: Fixed the output of list artifacts API calls  https://review.openstack.org/21460007:44
*** takedakn has quit IRC07:48
*** aix has joined #openstack-glance07:50
*** takedakn has joined #openstack-glance07:54
*** markus_z has joined #openstack-glance07:56
*** gberginc has quit IRC07:58
*** haomaiwang has quit IRC08:01
*** haomaiwang has joined #openstack-glance08:01
*** sayali has joined #openstack-glance08:04
*** itisha has joined #openstack-glance08:14
*** DuncanT has joined #openstack-glance08:18
*** gberginc has joined #openstack-glance08:19
*** sayali has quit IRC08:23
*** sayali has joined #openstack-glance08:27
*** takedakn has quit IRC08:27
*** MattMan has quit IRC08:43
*** MattMan has joined #openstack-glance08:44
*** aix has quit IRC08:46
*** haomaiwang has quit IRC09:01
*** haomaiwang has joined #openstack-glance09:01
*** aix has joined #openstack-glance09:16
openstackgerritOpenStack Proposal Bot proposed openstack/glance: Updated from global requirements  https://review.openstack.org/22461009:17
*** boris-42 has joined #openstack-glance09:26
*** e0ne has joined #openstack-glance09:26
*** dims has joined #openstack-glance09:47
*** marcusvrn_ has joined #openstack-glance09:49
*** ameade has joined #openstack-glance09:53
*** serverascode has joined #openstack-glance09:54
*** zhiyan has joined #openstack-glance09:54
*** haomaiwang has quit IRC10:01
*** haomaiwang has joined #openstack-glance10:01
*** niedbalski has quit IRC10:09
*** itisha has quit IRC10:23
*** e0ne has quit IRC10:30
*** e0ne has joined #openstack-glance10:35
*** gberginc has quit IRC10:36
*** smatzek has joined #openstack-glance10:37
*** itisha has joined #openstack-glance10:43
*** aix has quit IRC10:53
*** haomaiwang has quit IRC11:01
*** haomaiwang has joined #openstack-glance11:01
*** gberginc has joined #openstack-glance11:02
*** aix has joined #openstack-glance11:05
*** gberginc has quit IRC11:26
*** lakshmiS has quit IRC11:35
*** julim has joined #openstack-glance11:37
*** haomaiwang has quit IRC11:55
*** dosaboy has quit IRC12:00
*** e0ne has quit IRC12:03
ekarlsoa/buf4212:04
*** dosaboy has joined #openstack-glance12:07
*** e0ne has joined #openstack-glance12:23
openstackgerritKairat Kushaev proposed openstack/glance: Validate empty location value for v1 api  https://review.openstack.org/22629512:25
openstackgerritDarja Shakhray proposed openstack/glance: Fix 500 status code when we add in "depend_on" yourself  https://review.openstack.org/22058712:26
*** burgerk has joined #openstack-glance12:31
*** edmondsw has joined #openstack-glance12:33
buntingflaper87: ping12:35
flaper87bunting: pong12:36
buntingHey, i was just wondering your thoughts https://review.openstack.org/#/c/209051/ on how we could go about solving this problem12:38
bunting(not requiring the user to upload the whole image, before it tells them it has failed)12:38
*** achanda has joined #openstack-glance12:41
openstackgerritKairat Kushaev proposed openstack/python-glanceclient: Replace exception_to_str with oslo.utils function  https://review.openstack.org/22630312:41
*** achanda has quit IRC12:42
openstackgerritDarja Shakhray proposed openstack/glance: Fix 500 error when we specify invalid headers when work with blob/bloblist  https://review.openstack.org/22105612:45
*** alejandrito has joined #openstack-glance12:47
*** boris-42 has quit IRC12:49
*** haomaiwang has joined #openstack-glance13:00
*** haomaiwang has quit IRC13:01
*** haomaiwang has joined #openstack-glance13:01
*** MattMan has quit IRC13:12
*** MattMan has joined #openstack-glance13:13
*** dims has quit IRC13:14
*** dims has joined #openstack-glance13:15
*** abhishekk has quit IRC13:22
*** jecarey has joined #openstack-glance13:29
*** burgerk has quit IRC13:31
*** smatzek has quit IRC13:34
openstackgerritNikhil Komawar proposed openstack/glance: Fix for Image members not generating notifications  https://review.openstack.org/22130713:35
openstackgerritLakshmi N Sampath proposed openstack/glance: Fix for Image members not generating notifications  https://review.openstack.org/22130713:36
*** e0ne has quit IRC13:37
*** e0ne has joined #openstack-glance13:39
*** e0ne has quit IRC13:39
*** e0ne has joined #openstack-glance13:40
*** takedakn has joined #openstack-glance13:45
*** ducttape_ has joined #openstack-glance13:53
*** cdelatte has joined #openstack-glance13:54
*** delattec has joined #openstack-glance13:54
openstackgerritDarja Shakhray proposed openstack/glance: Fix glance ignored a headers when created artifact  https://review.openstack.org/22047613:55
*** haomaiwang has quit IRC14:01
nikhilCourtesy Glance Drivers' meeting reminder: nikhil_k, flaper87, sigmavirus24, rosmaita14:01
*** smatzek has joined #openstack-glance14:01
*** haomaiwa_ has joined #openstack-glance14:01
*** sgotliv has quit IRC14:03
*** stevemar has joined #openstack-glance14:03
*** burgerk has joined #openstack-glance14:05
openstackgerritFlavio Percoco proposed openstack/glance: Prevent image status being directly modified via v1  https://review.openstack.org/22633614:09
*** cdelatte has quit IRC14:12
*** delattec has quit IRC14:12
*** exploreshaifali has joined #openstack-glance14:14
*** ayoung has joined #openstack-glance14:16
*** vijendar has joined #openstack-glance14:20
*** ducttape_ has quit IRC14:25
*** harshs has joined #openstack-glance14:29
*** boris-42 has joined #openstack-glance14:30
nikhilo/14:31
dhellmanno/14:31
rosmaitao/14:31
*** cdelatte has joined #openstack-glance14:31
rosmaitadhellmann: we really did think through a lot of this already14:31
rosmaitai guess the key mistake was no API WG yet14:31
rosmaitapre def core14:31
rosmaitaand making the 'input' element a blob14:31
flaper87o/14:31
dhellmannrosmaita: yes! I hope I haven't given the impression that I think otherwise. I really do think the issue is communication, into and out of the team, rather than with the work per se.14:32
*** ayoung has quit IRC14:32
rosmaitadhellmann: cool14:32
dhellmannhaving a clear spec, describing the use cases and which API(s) meet each of the requirements, would be a good first step at communicating out of the team14:32
nikhilagreed14:32
nikhilconsolidation of the use cases too bar communication14:32
dhellmannand, as you say, there are some requirements that are new since the work started, and those may require changes/additions14:32
rosmaitaand i have been thinking about the question you posed about why the 'input' must differ from deployment to deployment14:32
rosmaitadhellmann: ^14:33
rosmaitaand i am not sure i have a completely convincing answer14:33
nikhildhellmann I echoed your words, thanks for stating that explicitly14:33
dhellmannnikhil: ++14:33
*** cdelatte has quit IRC14:33
nikhilone thing that might be relevant is14:34
dhellmannrosmaita: I can certainly understand why allowing the inputs to be different can make some things simpler for implementation, and even deployment.14:34
nikhilif we are planning to support federation of clouds14:34
dhellmannrosmaita: it's less great for the cloud user, though :-)14:34
rosmaitadhellmann: i guess we were counting on excellent tooling14:34
nikhilthen the details of the image may change per sub-domain in the federation14:34
nikhildhellmann one question, I am bound to ask is14:35
dhellmannrosmaita: yes, I think defcore pushes the "usability boundary" back inside the project a bit14:35
rosmaitai think the primary consideration is that glance must be rock-solid in working with nova14:35
nikhilhow to we define interoperability and communicate to the users14:35
flaper87rosmaita: nikhil dhellmann I wrote down, on the etherpad, some requirements earlier today and I see mordred is adding more there. Would it be ok for you to keep collecting things there until there's enough to go to a spec?14:35
*** ducttape_ has joined #openstack-glance14:35
flaper87I'm happy to remove the rest of the contents from that etherpad14:35
flaper87it's not really relevant14:35
dhellmannnikhil: as with flavors, I don't think you need the images in 2 clouds to be exactly the same as long as it is possible for a user to find the image they want/need14:36
nikhildhellmann yeah, and the input element in the task provides a way for the user to define the specifications of that image14:37
dhellmannnikhil: that said, you're thinking along the right lines -- how can the glance API make it easy for a user of one cloud to put some of their workload in a federated cloud if that workload relies on a specific image (specific in terms of its contents, not UUID or name)14:37
nikhillike whether they choose not to have a pre-specified image format, container format etc14:37
dhellmannflaper87: an etherpad is a good way to brainstorm14:37
nikhilexactly dhellmann14:37
dhellmannnikhil: cool, so we just need to standardize the "input" elements so that the user can (a) know what values can be passed and (b) pass the same values to 2 clouds and get the expected results14:38
dhellmannnikhil: I'm not sure why the user should care about image format once an image is in glance. Is that something they have to choose?14:39
*** cdelatte has joined #openstack-glance14:39
nikhildhellmann I think that can be done as we have been debating on when the input validation needs to happen. I guess the next step is to find the use cases that are okay with such validation of that parm at the API level with correct BadRequest message back to users for consistent for each cloud.14:39
nikhildhellmann the import of an image can provide a way to convert that image into a specified format14:40
nikhilconversion bit needs that in the input element14:40
dhellmannnikhil: is it possible to separate the discussion of how images get into the cloud and how an image of the right format is selected to deploy to create a given instance? because ultimately it would be nice if the user didn't have to care about format at either of those points, and there are many ways to make that happen14:40
nikhilI guess the discoverability needs to have the series of operations that need to/will  happen on image when imported14:40
openstackgerritMerged openstack/glance: Fixed the output of list artifacts API calls  https://review.openstack.org/21460014:41
flaper87I'm still not fully convinced the tasks API is what we want to use for such an essential piece of Glance. The current task implementation doens't resolve the issue of the user being able to upload the image to glance14:41
dhellmannand I don't think image conversion needs to be part of the create API spec, since it can happen transparently14:41
openstackgerritMerged openstack/glance: Remove duplicate name attribute  https://review.openstack.org/21029914:41
*** mtanino has joined #openstack-glance14:41
flaper87I think we should first think how the image bits are going to get into the cloud14:41
dhellmannnikhil: why does the user care about what is going to happen to the image after they import it?14:41
flaper87We have a way to do that and we can't change it just like that because that would break backwards compatibilty14:41
nikhildhellmann that was one of my main concern if we need to standardize the API then we should open a standarsize task vs. making import standardized -- one that has less of such customizations14:42
flaper87On the other hand, we also want people to moe to the new way without breaking apps developed using the old way14:42
flaper87Therefore these 2 need to be consistent somehow14:42
dhellmannflaper87, nikhil: you should think about this as though there is no existing API to figure out the requirements and what a good API would look like. Then as a second step, figure out how far the current API is from that ideal.14:42
flaper87dhellmann: exactly!14:42
openstackgerritStuart McLaren proposed openstack/glance: Prevent image status being directly modified via v1  https://review.openstack.org/22635514:42
nikhildhellmann if the operator chooses to support multiple type of virtualization and  user wants to select one of those, then the image format helps14:43
dhellmannbecause both existing APIs definitely have issues, and while we can't drop them we don't want to lock ourselves into keeping them around forever14:43
nikhildhellmann but if that use case is still in demand from the SuperUser would be my next question14:43
*** exploreshaifali has quit IRC14:43
dhellmannnikhil: I should be able to upload an image of any format one time and then deploy it to any hypervisor transparently without having to know any of those details. I shouldn't have to tell the cloud how it is deployed.14:43
nikhildhellmann agreed, both have issues14:43
nikhildhellmann but a cloud may choose to bill you differently based on the format of the image14:44
nikhiland that format could affect the performance of the virt14:44
nikhilthe choise won't be on scheduling14:44
nikhilbut at the level of deciding what's most value for money for booting my instance14:44
dhellmannnikhil: "may choose to" or "does actually choose to"?14:45
nikhildhellmann may choose to14:45
dhellmannbecause it's terrible from a usability standpoint to make a user think about that at any point other than when they are making an instance14:45
*** GB21 has joined #openstack-glance14:45
*** GB21_ has joined #openstack-glance14:45
dhellmannok, I posit that no one would ever actually charge me differently to store a raw image or an OVA image except that one might be a different size than the other, but they *would* charge me differently to run on a server where a raw image is required or a VM where an OVA image is required14:46
nikhildhellmann I am not sure how we can/should expose this to users . the dashboard could hide such details from them14:46
nikhilagain, the is the exact reason I want it to be a non-defcore API14:46
rosmaitai see mclaren has arrived on flaper87's etherpad14:47
*** harshs has quit IRC14:47
nikhildhellmann in a hybird cloud setting this may be actually req14:47
flaper87rosmaita: :P14:47
dhellmannnikhil: I should not have to upload an image twice to support a hybrid cloud.14:47
nikhilsay pub cloud support virt of type A and private support virt of type B14:47
nikhildhellmann so, would you expect a behind the scenes image conversion and record creation for this use case?14:48
dhellmannnikhil: yes14:48
rosmaitaso that adds to boot time14:48
nikhilthat breakes the immutability clause of the images14:48
nikhilbreaks*14:48
dhellmannnikhil: perhaps "an image" should have a 1:N relationship with "the place where image contents are stored" and that place has a format14:48
nikhilthe image format, container format, size, metadata on the image can vary in that case14:49
nikhiland those are immutable attrs14:49
*** TravT has quit IRC14:49
dhellmannwhat metadata other than the formats would change?14:49
nikhilalso, is the user has digital signon the image then that would fail miserably14:49
nikhilif*14:50
nikhildhellmann say virtual size14:50
rosmaitadhellmann: this was a use case for artifacts, and artifact could "point to" multiple images that would otherwise be singular14:50
nikhilEXACTLY! Thank you rosmaita !14:51
*** tsekiyama has joined #openstack-glance14:51
dhellmannok. I think your mental model of what an image is differs from the mental model of users of the cloud.14:51
*** tsekiyam_ has joined #openstack-glance14:52
rosmaitadhellmann: very possibly14:52
dhellmannthere are clearly some existing requirements I don't understand, so having those documented together with the new ones will help me continue the discussion14:52
mordrednikhil: I agree with dhellmann14:52
nikhilsure14:52
mordred"may choose to bill you differently based on the format of the image" makes no sense to me as an end user14:52
mordredas an end user, an image is a bundle of a filesystem, and if a cloud wants it in a different format, I'd like for them to do that magically without me knowing14:53
dhellmannright14:53
nikhiland that for image model is creating a new image14:53
mordredI conver images to vhd for rackspace not because I've made a decisoin bout format, but because I have no choice14:53
nikhilwhat image conversion can do for you14:53
*** gberginc has joined #openstack-glance14:54
dhellmannnikhil: right, so it's limited today, and that's fine. Let's design an API that allows us to remove that limitation in the future.14:54
mordred++14:54
nikhilmordred sure, and not all virtualization standards will support all image formats14:54
dhellmannwhether that is the artifacts API or something else, I don't know14:54
mordredif I uploaded a qcow2 image and the image was converted to vhd as it was stored14:54
nikhildhellmann sure14:55
*** ayoung has joined #openstack-glance14:55
mordredI would not, as a user, consider that my image had been modified14:55
flaper87mordred: ++14:55
rosmaitamordred: but some users care about that a lot14:55
nikhilmordred well, it really depends on the user14:55
mordredreally?14:55
rosmaitaCERN cares a lot14:55
nikhilfor a user who care about the immutability of image data it matters14:55
mordredwhy?14:55
nikhilso do other research labs14:55
rosmaitaall the image sig stuff indicates that people care14:55
jokke_mordred: how do you verify that as your own tools and your cloud tools provide you different checksum on that image?14:56
nikhilyep14:56
*** tsekiyama has quit IRC14:56
dhellmannjokke_: by looking at the contents of the image in a format-independent way14:56
mordredcan you explain what it is that they care about?14:56
nikhildhellmann and that will slow down boot times14:56
dhellmannI can make a tarball and a zipfile of the same directories and get a different checksum, but the contents are the same14:56
mordreddhellmann: ++14:56
nikhilthe question is who verifies that14:57
rosmaitamordred: with CERN they need to know that the distributed computations are done by identical systems with absolutely no differences14:57
nikhilI doubt if nova would like to do that at the pre-boot process14:57
dhellmannnikhil: then maybe it shouldn't be done at boot time. Maybe I should be able to sign an image myself, to know that the checksum of the format it's being kept in has not manipulated the contents14:57
mordredok. so - let me make a suggestion14:57
flaper87(folks, I'm sorry but I have to step out, I'll read the logs and the etherpad)14:57
mordredthe cloud should be able to tell me what image formats I can upload an image in14:57
rosmaitamordred: that i agree with entirely14:58
*** nikhil has quit IRC14:58
jokke_mordred: that I do agree14:58
dhellmannthere you go, if the cloud is going to convert to some format and I can upload in that format then I can do the checksum myself14:58
mordredit should also tell me what image format it prefers14:58
rosmaitawe have left that to documentation14:58
*** nikhil has joined #openstack-glance14:58
mordredand it should let me upload an image and say "please convert this to the right format for me"14:58
mordredor to upload it in the preferred format myself14:58
mordredyeah?14:58
dhellmann so users who don't care about signatures can choose any format and users who do care can chose the default format14:58
dhellmannmordred: ++14:58
mordredyup14:58
mordredbecause _I_ would love to not have to run a patched version of vhd-util in my image create process. BUT - it makes total sense from what you say that other people may prefer to do that rather than have their cloud modify their upload14:59
*** spzala has joined #openstack-glance14:59
dhellmannright, so that's a new API endpoint to get the list of supported image formats, possibly with details like links to documentation for them, and some indication of which is the preferred for a given region, zone, flavor, etc.15:00
mordredand I'd like to argue for making is simple for all clouds ot accept qcow2 as a format that they will have an opt-in convert-from thing for, but I don't need to die on a hill for it15:01
*** haomaiwa_ has quit IRC15:01
*** GB21_ has quit IRC15:01
jokke_The preferred format would also make sense when we go to signed or encrypted images. If the user wants to make sure their image does not change between upload and boot, they should have clear indication in what format they are expected to have the image15:01
mordred(this is largely because qcow2 is ubiquitous in terms of making and also smaller than the other formats so there's less network transit)15:01
mordredjokke_: ++15:01
*** ducttape_ has quit IRC15:01
dhellmannis someone capturing this in the etherpad?15:02
jokke_if not, this channel is logged15:02
*** TravT has joined #openstack-glance15:02
dhellmannjokke_: well, yeah, but we're putting stuff in this etherpad so it's in one place15:02
*** haomaiwa_ has joined #openstack-glance15:02
dhellmannI didn't want to duplicate it and I haven't been following the editing15:02
mordredfor context - the ubuntu-trusty image that infra uploads as the base for devstack jobs is 5.8G in qcow2 and 16G in both raw and VHD format15:03
* nikhil back after irc troubles...15:03
rosmaitamordred: what's the compression method with acow2 ?15:03
mordredwhen you combine that with 4 different base images and the fact that we upload daily, the difference in size is non-trivial15:03
mordredrosmaita: I have no idea. it's just whatever spits out of qemu-img15:04
rosmaitaok15:04
rosmaitaso OVA would be good, but glance currently isn't architected to handle it (too dangerous)15:05
rosmaitaall the tgz attacks15:05
mordrednod15:05
jokke_dhellmann: my point was if no-one is capturing this real time it's not gonna be lost15:05
nikhilso, I think for a case where we have standard format for an image and then there are versioned formats for that image say qcow2 , vhd, etc15:05
mordredI mean, end of the day, if the format everythin accepted was RAW - it would still be an improvement because I wouldn't have 3 different copies :)15:05
*** takedakn has quit IRC15:05
dhellmannjokke_: ok15:06
nikhilmordred right and in that case, it makes sense to version that image15:06
mordrednikhil: I'm not sure I follow the concept of versioning an image?15:06
nikhilmy image foo has version foo-qcow2 (Standard), foo-vhd (version 2), foo-raw(version 3) ...15:06
dhellmannnikhil: as a user, I do not care about those differences once the image is in the cloud15:07
jokke_rosmaita: qcow2 uses zlib for compression15:07
nikhilsure, but if we decide on a single standard format of an image15:07
dhellmannnikhil: we are not doing that, though15:08
dhellmannwe're saying that a *given* cloud might have a *default*15:08
dhellmannand that should be discoverable, specifically to handle the signed image use case15:08
jokke_++15:08
dhellmannbut that it should not be *required* for folks who don't care about signed images15:08
nikhilso, for a federated cloud that can't be consitent for priv and public15:08
rosmaitajokke_: ty15:09
mordredwell. I'm arguing for a single format that all clouds would accept as an input into the "convert it for me" case15:09
dhellmannnikhil: it will not be the same, but because it is discoverable in a consistent way tools can deal with it15:09
dhellmannmordred: oh, I just thought we'd want any-to-any conversion15:09
mordredwhich is "here is an image, I do not care what you do with it, I just want its contents there where I boot it"15:09
mordreddhellmann: I DO want that15:09
dhellmannbut sure, if we want to set a minimum for that, that's a good start15:09
nikhilmordred yep, so I was saying that everyone needs to upload into this standard raw format15:09
dhellmannnikhil: no!15:09
mordrednikhil: but I thnk that's a hardship for the CERN case15:09
nikhiland version that image to an acceptable format of your deployment15:09
dhellmannthe cloud must support upload and conversion in that format, but users must not be required to use that format15:10
mordreddhellmann: right15:10
nikhilthe clouds would use what's suitable for the HV15:10
mordredit's a convenience baseline that people should alwyas be able to count on, even if it's inefficient15:10
nikhildhellmann So, you are saying that we do not enforce this for the image-signing?15:10
mordredyes15:10
dhellmannnikhil: if the upload process gives me a UUID as an identifier for the image I am creating, then I need to be able to use *that* UUID when booting an instance, not a different UUID for the "right version" of an image15:10
rosmaitaso ... if you try to import a windows image to AWS, they install some licensing stuff on it ... do we foresee anything like that happening in an openstack deployment?15:11
mordredrosmaita: wouldn't it be friendlier to have the cloud expose licensing stuff into config-drive and let windows-cloud-init consume it?15:11
dhellmannnikhil: "be liberal in what you accept" -- allow the user to upload in (a) a least-common-denominator format (b) the preferred format for a given cloud (c) any other format -- and for cases where (a) and (c) do not match (b) the cloud should convert for me15:11
dhellmannwhere the any-to-any conversion required by (c) can be done in a later cycle15:12
mordredby that I mean, I _never_ want the cloud to modify my content (which is why file injection was such a bad idea)15:12
dhellmannactually, all of this image conversion can be done later, but...15:12
nikhilbut that simply breaks the immutabiliy clause15:12
*** harshs has joined #openstack-glance15:12
dhellmannnikhil: option (b) gives you immutability15:12
nikhilyes15:13
rosmaitamordred: i kind of agree, but we also want things to "just work" for users15:13
dhellmannnikhil: many of your users care far far less about your definition of immutability, and will happily use (a) and (c)15:13
mordredrosmaita: indeed. I have less of a person background in windows images15:13
mordredrosmaita: so I don't have a good pov on which thing is better/worse15:13
dhellmannnikhil: so now you're giving your users a choice between convenience [(a) and (c)] and hard-line immutability (b)15:13
nikhilso, that the tradeoff we need to hit15:13
mordredrosmaita: I just know that in linux file injection makes things break15:13
mordredrosmaita: because my cloud will helpfuly overwrite things my config management system has put there :)15:14
dhellmannI propose that we concentrate on (b) for this cycle, with an eye on making sure we don't make it impossible to have (a)15:14
*** annegentle has joined #openstack-glance15:14
mordredand "injectinting some license stuff" reminds me of that - but I'm not sure what's the norm in windows admin land15:14
rosmaitadhellmann: that sounds reasonable15:14
mordredrosmaita: you know ...15:14
rosmaitamordred: i'm not completely sure either15:14
dhellmannmordred: do you agree with deferring image conversion for now?15:15
nikhildhellmann does it matter what our SuperUsers want? It would be nice to get some feedback from users on that immutability clause. number of users vs. strong adopters15:15
mordreddhellmann: do you mean deferring the decision?15:15
*** cdelatte has quit IRC15:15
dhellmannmordred: no, deferring supporting that as a feature as long as the API doesn't preclude it15:15
mordreddhellmann: yes. I agre with b15:16
rosmaitai think end-users don't care about immutability until they realize that it's important15:16
rosmaitaanyway, b sounds reasonable to me, too15:16
nikhilyeah15:16
dhellmannnikhil: I think we're talking past each other. I have described how we can support immutability if that is what the user wants, and convenience if that is what they want.15:16
mordredyah. also - for different things15:16
rosmaitaok, next question is the actual upload15:16
mordredI do not care about immutability for nodepool images at all - I would care more about immutable/signed images if I uploaded base images for our control plane15:16
dhellmannright, so it's an image-by-image concern rather than all or nothing15:17
nikhilagreed15:17
mordred(although to be clear - I think we should break immutability into two facets - immutability of content and immuntability of form)15:17
*** TravT has quit IRC15:17
dhellmannmordred: yes, different levels of guarantee makes sense15:17
rosmaitamordred: that's conceptually clear, but can it be done in practice?15:18
mordredI care about immutability of content everywhere, (no file injection) but I recognize it's hard to trust a cloud-provided guarantee of that if I have allowed the cloud to mutate form15:18
nikhiland should it be done in glance too is my question15:18
mordredrosmaita: no. it's not provable in practice15:18
rosmaitaexactly15:18
dhellmannnikhil: I'm not sure what you mean?15:18
mordredrosmaita: but - if I cannot trust my cloud to not lie to me15:18
mordredrosmaita: then I'm screwed15:18
mordredrosmaita: becauseyou could store my checksum when I upload and then boot something else without telling me15:19
rosmaitawell, not really15:19
rosmaitathere is a workaround15:19
nikhilI know for large scale boot processed non-immutability (mutating image behind download transactions) can be very dangerous and result into waste of cloud resources15:19
mordredrosmaita: my cloud provider has root access to the machines my vms are running on15:19
nikhilprocesse*15:19
mordredat a certain point, I have to trust them somewhat15:19
*** dims has quit IRC15:20
nikhildhellmann I was asking about the faceting of immutability concerns ...15:20
mordrednikhil: sorry15:20
mordrednikhil: I was mainly mentoining that in terms of our conversations about the topic15:21
mordrednot in terms of a concept that should be exposed pe se15:21
mordredper se15:21
nikhilAh I see :)15:21
*** mclaren has joined #openstack-glance15:21
* nikhil is listening15:21
mclarenjust lurking :-)15:21
openstackgerritLakshmi N Sampath proposed openstack/glance: Fix for Image members not generating notifications  https://review.openstack.org/22130715:21
dhellmannthis has been a really good conversation, but unfortunately I have to go to a meeting and then do some release things15:21
* nikhil iswith some n/w lag now...15:22
*** tsekiyam_ has quit IRC15:22
dhellmannnikhil, rosmaita, flaper87, jokke_, mordred : thanks, I think we're all starting to understand the multiple perspectives better as a result of talking today15:22
nikhilmordred yeah, that makes sense. And Artifacts is what I think is the solution to this problem15:22
mordreddhellmann: ++15:22
rosmaitadhellmann: cool, next thing we need to settle is the upload parameters15:22
rosmaitai mean what is acceptable15:23
*** tsekiyama has joined #openstack-glance15:23
mordredrosmaita: ++15:23
nikhiland it's another big conversation topic15:23
nikhildhellmann ++15:23
mclarenmissed the party, sorry15:23
rosmaitamclaren: cheers15:23
nikhilrosmaita I think that would depend on the winning use cases afait15:24
mordredmclaren: I'm sure there will be more parties15:24
dhellmannmclaren: the scrollback is worth a read-through15:24
mclarensure15:24
rosmaitamclaren: conversations started in this channel at 15:00 UTC15:24
mclarenack15:24
nikhilTo me the puzzle seems to be of two main divisions -- 1. Number of users who do not care about certain clauses viz. immutability, upload mechanism etc 2. SuperUsers who has certain specific requirements to each ofthose. Would be good to settle in on such.15:25
nikhilhave*15:25
nikhilmclaren there's some more from driver's meeting15:26
mclarennikhil: ok, plenty of bedtime reading15:26
mordrednikhil: agree. the balance between meeting the needs of the people who do care and not burdening the people who do not care is key15:26
nikhilmordred well put! cloud be today's status :-)15:27
nikhilha15:27
nikhilI meant could be15:27
*** annegent_ has joined #openstack-glance15:27
* nikhil running to wind up Liberty ...15:27
* nikhil thanks all for the sync15:28
*** e0ne has quit IRC15:28
*** dims has joined #openstack-glance15:28
*** harshs has quit IRC15:28
*** e0ne has joined #openstack-glance15:30
*** MattMan has left #openstack-glance15:30
*** annegentle has quit IRC15:30
*** TravT has joined #openstack-glance15:36
*** annegent_ has quit IRC15:36
*** annegentle has joined #openstack-glance15:36
*** lakshmiS has joined #openstack-glance15:39
*** cdelatte has joined #openstack-glance15:39
*** cdelatte has quit IRC15:40
*** gberginc has quit IRC15:41
*** e0ne has quit IRC15:41
*** TravT has quit IRC15:46
*** TravT has joined #openstack-glance15:48
*** gberginc has joined #openstack-glance15:49
*** cdelatte has joined #openstack-glance15:50
*** cdelatte has quit IRC15:54
*** f13o has joined #openstack-glance15:55
*** cdelatte has joined #openstack-glance16:00
*** 1JTAAB3VD has joined #openstack-glance16:00
*** cdelatte has quit IRC16:00
*** 1JTAAB3VD has quit IRC16:00
*** TravT has quit IRC16:00
*** haomaiwa_ has quit IRC16:01
*** haomaiwa_ has joined #openstack-glance16:01
*** kebray has joined #openstack-glance16:03
*** whydidyoustealmy is now known as shakamunyi16:03
*** kebray has joined #openstack-glance16:03
*** ducttape_ has joined #openstack-glance16:06
*** annegent_ has joined #openstack-glance16:06
*** annegentle has quit IRC16:07
*** exploreshaifali has joined #openstack-glance16:12
*** ducttape_ has quit IRC16:13
*** cdelatte has joined #openstack-glance16:18
*** TravT has joined #openstack-glance16:24
*** dosaboy_ has joined #openstack-glance16:25
*** dosaboy_ has quit IRC16:26
*** ducttape_ has joined #openstack-glance16:27
*** TravT has quit IRC16:28
*** ducttape_ has quit IRC16:32
*** dosaboy has quit IRC16:32
*** dosaboy has joined #openstack-glance16:33
*** ducttape_ has joined #openstack-glance16:37
*** annegent_ has quit IRC16:40
*** groen692 has quit IRC16:41
*** annegent_ has joined #openstack-glance16:42
*** TravT has joined #openstack-glance16:42
*** harshs has joined #openstack-glance16:46
*** david-lyle has quit IRC16:46
*** achanda has joined #openstack-glance16:47
*** annegent_ has quit IRC16:48
*** delattec has joined #openstack-glance16:49
*** ducttape_ has quit IRC16:50
*** cdelatte has quit IRC16:52
*** delattec has quit IRC16:53
*** markus_z has quit IRC16:54
*** haomaiwa_ has quit IRC17:01
*** haomaiwang has joined #openstack-glance17:01
*** delattec has joined #openstack-glance17:02
*** ducttape_ has joined #openstack-glance17:06
*** davideagnello has quit IRC17:08
*** kebray has quit IRC17:09
*** davideagnello has joined #openstack-glance17:10
*** gberginc has quit IRC17:13
*** achanda_ has joined #openstack-glance17:17
*** achanda has quit IRC17:17
*** gberginc has joined #openstack-glance17:20
*** kebray has joined #openstack-glance17:22
*** kebray has quit IRC17:26
*** dims has quit IRC17:31
*** dims has joined #openstack-glance17:32
*** gberginc has quit IRC17:36
*** haomaiwang has quit IRC17:39
*** delattec has quit IRC17:44
openstackgerritLakshmi N Sampath proposed openstack/glance: Fix for Image members not generating notifications  https://review.openstack.org/22130717:55
*** ducttape_ has quit IRC17:57
*** exploreshaifali has quit IRC17:57
*** delattec has joined #openstack-glance18:11
*** gberginc has joined #openstack-glance18:15
*** GB21 has quit IRC18:18
*** delattec has quit IRC18:21
openstackgerritVenkatesh Sampath proposed openstack/glance: Fix server start ping timeout for functional tests  https://review.openstack.org/22539118:23
*** achanda has joined #openstack-glance18:23
*** achanda_ has quit IRC18:24
*** ducttape_ has joined #openstack-glance18:25
openstackgerritMerged openstack/glance: Prevent image status being directly modified via v1  https://review.openstack.org/22633618:34
flaper87o/18:36
flaper87hey folks, sorry for stepping out earlier. Did I miss all the fun?18:36
sabariflaper87 Looks like the party's over :)18:38
flaper87sabari: :( sad panda18:38
sabarihaha18:39
*** ayoung has quit IRC18:39
*** achanda has quit IRC18:42
*** ducttape_ has quit IRC18:46
*** achanda has joined #openstack-glance18:48
*** aix has quit IRC18:49
*** ducttape_ has joined #openstack-glance18:50
*** ayoung has joined #openstack-glance18:54
*** ducttape_ has quit IRC18:58
*** 6JTAB3HZ2 has joined #openstack-glance19:01
*** 16WAAB3T7 has joined #openstack-glance19:01
openstackgerritGiridhar Jayavelu proposed openstack/glance_store: VMware: Fix missing space in error message  https://review.openstack.org/22650619:04
*** kebray has joined #openstack-glance19:07
*** kebray has quit IRC19:08
*** kebray has joined #openstack-glance19:08
openstackgerritLakshmi N Sampath proposed openstack/glance: Fix for Image members not generating notifications  https://review.openstack.org/22130719:09
*** e0ne has joined #openstack-glance19:10
*** exploreshaifali has joined #openstack-glance19:14
*** achanda has quit IRC19:37
*** 16WAAB3T7 has quit IRC19:47
*** 6JTAB3HZ2 has quit IRC19:47
*** delattec has joined #openstack-glance19:49
*** delatte has joined #openstack-glance19:49
*** ducttape_ has joined #openstack-glance19:51
*** itisha has quit IRC19:53
*** kebray has quit IRC19:57
*** kebray has joined #openstack-glance19:58
*** annegentle has joined #openstack-glance20:02
*** jamielennox|away is now known as jamielennox20:03
*** jecarey has left #openstack-glance20:09
*** achanda has joined #openstack-glance20:15
*** smatzek has quit IRC20:22
*** rosmaita has quit IRC20:29
*** rosmaita has joined #openstack-glance20:33
*** lakshmiS has quit IRC20:37
*** ducttape_ has quit IRC20:39
*** ducttape_ has joined #openstack-glance20:41
*** ducttape_ has quit IRC20:42
*** TravT has quit IRC20:45
*** _ducttape_ has joined #openstack-glance20:47
*** _ducttape_ has quit IRC20:50
*** e0ne has quit IRC20:51
*** annegentle has quit IRC20:54
*** ducttape_ has joined #openstack-glance20:54
*** ducttape_ has quit IRC20:58
*** ayoung has quit IRC21:09
*** kebray has quit IRC21:21
*** achanda has quit IRC21:21
*** ayoung has joined #openstack-glance21:22
*** annegentle has joined #openstack-glance21:24
*** ayoung has quit IRC21:29
*** alejandrito has quit IRC21:32
*** alejandrito has joined #openstack-glance21:33
*** ducttape_ has joined #openstack-glance21:35
*** ducttape_ has quit IRC21:35
*** ducttape_ has joined #openstack-glance21:36
*** delatte has quit IRC21:37
*** delattec has quit IRC21:37
*** delattec has joined #openstack-glance21:39
*** delatte has joined #openstack-glance21:39
*** burgerk has quit IRC21:42
*** TravT has joined #openstack-glance21:50
*** annegentle has quit IRC21:54
*** annegentle has joined #openstack-glance21:54
*** delatte has quit IRC22:03
*** delattec has quit IRC22:03
*** ducttape_ has quit IRC22:07
*** delattec has joined #openstack-glance22:08
*** delatte has joined #openstack-glance22:08
*** edmondsw has quit IRC22:10
*** Guest61243 is now known as mfisch22:11
*** mfisch is now known as Guest5071522:12
*** vijendar has quit IRC22:13
*** Guest50715 is now known as mfisch22:13
*** mfisch has quit IRC22:13
*** mfisch has joined #openstack-glance22:13
*** boris-42 has quit IRC22:22
*** achanda has joined #openstack-glance22:22
*** boris-42 has joined #openstack-glance22:24
*** stevelle has quit IRC22:26
*** stevelle_ has joined #openstack-glance22:26
*** briancurtin has quit IRC22:26
*** achanda has quit IRC22:27
*** briancurtin has joined #openstack-glance22:28
*** frickler has quit IRC22:29
*** frickler has joined #openstack-glance22:29
*** spzala has quit IRC22:34
*** delatte has quit IRC22:42
*** delattec has quit IRC22:42
*** ayoung has joined #openstack-glance22:44
*** alejandrito has quit IRC22:55
*** annegentle has quit IRC22:56
*** ducttape_ has joined #openstack-glance22:58
*** ducttape_ has quit IRC23:00
*** ducttape_ has joined #openstack-glance23:01
*** delatte has joined #openstack-glance23:04
*** delattec has joined #openstack-glance23:04
*** spzala has joined #openstack-glance23:04
*** achanda has joined #openstack-glance23:12
*** ducttape_ has quit IRC23:14
*** spzala has quit IRC23:18
*** dims has quit IRC23:18
*** stevemar has quit IRC23:23
*** stevemar has joined #openstack-glance23:32
*** stevemar has quit IRC23:36
sabariwhat's the reason to still have the sample conf s in the source? nikhil jokke_23:42
sabariI keep forgetting - I knew there was some reason for it.23:42
nikhili think devstack uses it23:42
nikhil also may be other small deployments23:43
sabariah thanks Nikhi !23:43
sabarithough not sure if I understand the small deployments case.23:44
*** dims has joined #openstack-glance23:54
*** dims has quit IRC23:55
*** Guest18166 has joined #openstack-glance23:55

Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!