14:00:32 #startmeeting glance 14:00:33 Meeting started Thu Mar 29 14:00:32 2018 UTC and is due to finish in 60 minutes. The chair is jokke_. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:00:34 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 14:00:36 The meeting name has been set to 'glance' 14:00:40 #topic roll call 14:00:45 o/ 14:00:45 o/ 14:00:50 o/ 14:00:53 smcginnis too 14:01:00 #link https://etherpad.openstack.org/p/glance-team-meeting-agenda 14:01:29 o/ o/ :) 14:01:43 \o 14:01:58 o/ 14:02:16 #topic updates 14:03:24 so pretty muc one update ... trying to find the mail 14:04:08 but in nutshell we're moving from maintaining the minimum requirements centrally to in-repo requirement management 14:05:19 and there will be similar lower-constrains job as there is upper constrains 14:05:45 http://lists.openstack.org/pipermail/openstack-dev/2018-March/128352.html 14:05:51 Saw that patch to be reviewed 14:05:54 for the lower constraints 14:05:59 https://review.openstack.org/#/c/555483/ 14:06:28 dhellmann did script the needed changes, so I'd like everyone to get familiar with how it affects us, and what pitfall we need to look out for 14:06:35 rosmaita: thanks 14:06:41 np 14:07:00 #link http://lists.openstack.org/pipermail/openstack-dev/2018-March/128352.html 14:07:26 that's a fairly long email! 14:07:49 other update is that we still haven't got the glanceclient released so lets hop into topic 2 14:08:09 #topic python-glanceclient image-import default behavior 14:08:36 rosmaita: you wanna start on this? 14:08:50 sure 14:08:52 #link https://review.openstack.org/#/c/555550/ 14:09:08 ok, so i was testing out glanceclient and came across some situations 14:09:29 decided to handle them by doing more analysis on the client side to see if a request was likely to succeed before making it 14:09:47 because the problem is that you can have client-side import succes, that is the import call succeeds 14:09:57 but hte actual import fails later when the task is processed 14:10:02 and that can be problematic 14:10:03 anyway 14:10:35 as part of that, i changed the default import-method to none, and in that case all the image-create-via-import does is create an image record 14:10:50 or exits if you have included extra junk 14:11:02 so, the question is whether that is a good change 14:11:22 my reasoning is that we have >1 import method, so the user should pick 14:11:43 but, that is backward incompatable with the current image-create command 14:12:00 assuming that we will eventually replace image-create with image-create-via-import 14:12:15 but, it might be worth doing anyway 14:12:30 anyway, that's the basic idea 14:12:50 i personally can go either way, key thing is that we've got tests to make sure the input makes sense 14:13:14 but obviously, i think making the user say what import-method they are using is a good idea 14:13:25 since that's what my patch does 14:13:34 * rosmaita shuts up 14:13:50 so, I'm hugely afraid about the outcry if we replace the image create client call with something that is not backwards compatible ... on the same time as image-create-via-import is experimental, we can experiment with it, not rename it but rewrite the image-create with sensible defaults when the time comes to swap 14:15:09 that is our benefit of flagging it experimental in first place, and big thanks to rosmaita pointing out that we should not lift that flag just yet 14:15:40 i guess the point of a decision now is that the openstack client people or sdk people may look at our implementation as a reference 14:16:33 or maybe that's the point of having a non-backward-compat change in the experimental command 14:16:50 +1 14:17:05 i guess my point is that i don't like silent defaults with complex choices 14:17:17 "explicit is better than implicit" 14:17:21 i heard that somewhere 14:17:58 rosmaita: yeah, so it's flagged experimental so we can get it running, test it and gather feedback. Once we make what ever the call is how it's behaving and remove that experimental flag, we need to start to be really careful what we break 14:18:11 jokke_ agree 100% 14:18:53 my short-term question is do we like the current patch, or do i need to rewrite it? 14:22:08 so I'm on favor of having sensible defaults or at least ability to set them 14:22:28 just a reminder: stable-maint team reviews release requests on mondays and we plan to release from stable/queens 14:23:02 and this coming Monday is easter so I quess we need to poke tonyb anyways 14:23:26 does RH have monday off? 14:23:42 I have both tomorrow and Monday off 14:23:49 slackers! 14:24:09 I have tomorrow off 14:24:44 i would suggest that a sensible default would do minimal damage, so in this instance, creating an image record only is a good choice 14:25:06 but that's just me 14:25:17 well we can still do that if the payload is not provided 14:25:55 but sensible default would be something that maintains backwards compatibility to established command 14:26:07 that's why it's there in the first place ;) 14:26:09 i guess my concern there is if we have multiple methods that take --file, that's a problem 14:26:18 maybe --uri is a better example 14:26:29 already a proposal for another method for that 14:27:23 yeah, and the default that is in place is backwards compatible and should not accept --uri 14:27:53 but like said, I'd be fine giving the user opportunity to set the default locally as well 14:27:55 well, let's take a vote or something ... i can make the changes right away and update all the tests if necessary ... looks like it needs to be done in the next few hours if we're going to propose a release for monday 14:28:19 having env variable for it, just like we have been doing with the api version 14:29:29 that's a good idea 14:29:35 I like that idea actually 14:29:44 and you mentioned it before but i ignored it 14:29:56 #startvote image-create-via-import should have sensible defaults? Yes, No, users-choise 14:29:57 Begin voting on: image-create-via-import should have sensible defaults? Valid vote options are Yes, No, users-choise. 14:29:58 Vote using '#vote OPTION'. Only your last vote counts. 14:30:13 crap 14:30:14 ummm ... we need to say what "sensible defaults" are! 14:30:17 #endvote 14:30:17 Voted on "image-create-via-import should have sensible defaults?" Results are 14:30:23 #startvote image-create-via-import should have sensible defaults? Yes, No, users-choice 14:30:23 Begin voting on: image-create-via-import should have sensible defaults? Valid vote options are Yes, No, users-choice. 14:30:25 Vote using '#vote OPTION'. Only your last vote counts. 14:30:57 #vote Yes 14:31:01 #vote Yes 14:31:03 rosmaita: something that does not break the world and get similar which hunt on us as the 1.0.0 release did :P 14:31:14 #vote Yes 14:31:17 #vote users-choice 14:31:40 #vote users-choice 14:32:34 ok, do we have anyone else here who has opinion? 14:33:37 User choice would be the environment variable right? The only thing I don't like about that is it's kind of a hidden thing that someone would need to go out and learn about. 14:34:20 yes, we would have to publicise it 14:34:26 smcginnis: yes, although I do not think it's a bad idea even if we keep the default in place to have that to obey 14:34:27 would show up in the help, i think 14:34:36 From a use case perspective I like sensible defaults 14:35:04 yes, it's like apple pie 14:35:15 Who doesn't like apple pie 14:35:16 not sure what the cross-cultural equivalent of that is 14:35:21 #endvote 14:35:22 Voted on "image-create-via-import should have sensible defaults?" Results are 14:35:23 users-choice (2): jokke_, rosmaita 14:35:25 Yes (3): McClymontS, smcginnis, abhishekk 14:35:26 If we can do user-choice, but have the no-env-variable-defined default be a reasonable one, then that's kind of the best of both voted options, right? 14:35:39 smcginnis i think so 14:35:41 User choice with reasonable default. 14:35:48 I'm good with that then. 14:35:49 rosmaita: you've seen movie American Pie, right? :D 14:35:51 yeah I think that meets in the middle 14:36:13 :D 14:36:20 ok, so can someone summarize what we agreed on? is the "sensible default" == glance-direct or None? 14:37:06 glance-direct would be the sensible backwards compatible default 14:37:13 I guess sensible default 14:38:14 ok, so i will put up a patch with glance-direct default and that recognizes the env var and revise the tests accordingly 14:38:22 unless you want to implement the magic into the client to detect what the user wants to do when they don't specify the method but provide payload ;) 14:38:43 well, that's basically what we 14:38:50 re doing 14:39:49 but now there is no way to simply create an image record 14:39:53 umm nope ... we are telling that _if_ you do image create and provide your image file, we will create the image all the way like we used to, and we have options for you to choose other methods ;) 14:40:15 because glance-direct requires data 14:40:28 so this is also backward incompatible 14:40:45 rosmaita: yes there is, we just haven't implemented it in the current implementation. So we do fail if the payload is not there, while we should just create the record as we used to 14:41:23 and yes, I was thinking of that when I looked the code and I have bug text pretty much written for it ;) 14:41:51 i am liking this less and less 14:42:19 ok so currently: 14:42:24 so, the import-method will have to remain None 14:42:42 but if you include --file or stdin we will know that you mean glance-direct and adjust 14:42:44 1) image create without payload - creates image that stay's in queued 14:43:07 we have to do it in the function, because the config-parser sets the default value 14:43:19 2) image create with payload - creates image and uploads the payload == if everything goes fine active image 14:43:39 with the import version where we should be: 14:44:02 1) image create without file - creates image that stays queues 14:44:48 2) image create with payload file or stdin - creates image, stages the payload, imports == if everything goes fine active image 14:45:06 with import and any other method choosen: 14:45:39 why don't we just drop the import-method altogether? 14:45:47 1) provide all the needed parameters for that method - creates image, does what is needed by the method, imports == if everything goes fine active imagew 14:46:55 I strongly dislike the image create to queued with no payload 14:47:46 McClymontS: there is a lot of legimite usecases to create image before providing the data and I'm not gonna let anyone break image-create around that 14:48:08 Understood, I understand why it was built out 14:48:45 Is it rate limited on default? 14:48:47 I'm not even sure 14:49:17 ok, so in the code, how do i distinguish between wanting to create an image record, and glance-direct where someone forgot the --file or stdin redirect? 14:49:33 McClymontS glance does not do rate limiting, that has to be done externally 14:49:37 McClymontS: I hear what you say and I'm inclined to say that image-create should have never had the option to "magically" detect what the user might want, but I did not have a say to that decision 14:50:16 Cool thanks for the context, rosmaita I'm not sure tbh 14:50:34 I think the best case scenario (and where rosmaita seems to want to move) would have been to tell the user, "This workflow needs to you run these commands and provide them their respective imput" 14:50:50 instead of "Dump everything to that first command and voodoo happens" 14:50:54 you have detected my hidden agenda! 14:51:28 rosmaita: and that hidden agenda is backwards incompatibe and I still remember there version 1.0.0 release way too well 14:51:57 ok, so i still have my question above about how can i tell in the function 14:52:50 what i mean is this: https://review.openstack.org/#/c/555550/2/glanceclient/v2/shell.py@105 14:52:54 line 105 14:53:42 if default='glance-direct', args.import_method will be == 'glance-direct' when control hits the function 14:54:28 the problem is that this is a more complicated command than image-create was 14:55:05 i mean, maybe the solution is just that we leave image-create as-is 14:55:09 into the futurew 14:55:12 rosmaita: so your worry is that there is no way to tell if user explicitely called it with method glance-direct and forgot to provide the payload? 14:55:18 exactly 14:55:25 we want to minimize support calls 14:55:39 i am thinking that this needs to be its own call 14:56:03 its own command, i mean 14:56:50 4 minutes 14:56:57 yeah, sorry about the time 14:56:59 ok, so I'm fone changing that defualt to None for that purpose and then we just need to specify that _if_ the method is not provided and payload is for backwards compatibility we default to glance-direct 14:57:08 rosmaita: Meeting hog. :) 14:57:16 :%s/fone/fine/ 14:57:29 yeah, but the help text is going to be weird 14:57:37 and the we need to do that detection in the code, which is very trivial 14:57:41 I know 14:58:02 but it's way less weird if you write it rather than me 14:58:07 :) 14:58:11 ok last 2min for open discussion 14:58:18 #topic open discussion 14:58:41 yeah, one last thing ... if this is a new call, there are no backward-compat worries 14:58:53 and operators are still going to want to use direct upload 14:59:06 so i really wonder about this being a 1-1 replacement for image-create command 14:59:17 rosmaita: well it's not because we promised this as drop in replacement for users (IIRC the spec right) 14:59:31 jokke_: waiting for new upstream branch for policy refactor changes :) 14:59:43 bhagyashris: it's created forgot it from updates 14:59:44 i don't remember the spec saying anything about client support 15:00:14 jokke_: ohh ok no problem 15:00:15 feature/policy-refactor 15:00:21 thanks all! 15:00:28 #endmeeting