Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2019-02-21
08:04:05 slaweq maybe it will not be very hard :)
09:13:55 frickler slaweq: oh, great, so this is only triggered by port-forwarding being enabled? and that scenario is missing in neutron gate?
09:15:24 slaweq frickler: to be strict it is problem with multiple loading L3 service plugin
09:15:49 slaweq because now it is added as "requirement" for port forwarding service plugin and is loaded from config too in Your scenario
09:15:52 slaweq and that cause this issue
09:16:24 slaweq I'm know what is wrong and now I'm trying to change the logic in such way that it will not init service plugin class twice
09:19:32 frickler slaweq: o.k., that sounds plausible, I'm still worried that the patch did pass neutron gate
09:19:54 slaweq frickler: yes, that is something else what I will have to check
09:20:03 slaweq frickler: but first I would like to fix the issue :)
09:53:07 frickler slaweq: sure. fyi I confirmed that your revert fixes the issue in my local devstack, but I'm assuming you are much further in your analysis anyway by now ;)
09:54:44 slaweq frickler: yes, I have locally patch which should fix this issue without reverting
09:54:56 slaweq frickler: I hope I will finish it in 1-2 hours and will send to review
10:37:03 openstackgerrit Jens Harbott (frickler) proposed openstack/openstacksdk master: DNM: Dummy patch to verify neutron fix https://review.openstack.org/638385
10:37:20 frickler slaweq: ^^ testing your patch here
10:38:19 slaweq frickler: ok, thx
10:39:41 slaweq frickler: I think I know why it didn't failed in neutron gate
10:40:07 slaweq frickler: we simply don't have any API/scenario tests for port forwarding and we are not configuring this service plugin in our jobs
10:40:24 slaweq frickler: and that is the only plugin which for now could trigger this issue :)
10:40:49 slaweq I will open another bug for that but I don't know exactly when I will have some time to get to it
10:43:15 frickler slaweq: maybe just running one of the sdk jobs in neutron as non-voting could be good enough. or at least until you get something proper into neutron-tempest-plugin scenarios
10:44:00 Shrews mordred: coolio
10:44:16 Shrews slaweq: thx for looking into that neutron issue
10:44:35 slaweq frickler: in last denver PTG we agreed that we can add some stadium projects' jobs to neutron queue to avoid such breaks
10:44:46 slaweq frickler: I think that adding SDK job there would be also good idea
10:45:00 slaweq as it is know it may have some value :)
10:45:13 slaweq frickler: should I do it or You will propose patch?
10:45:40 frickler slaweq: I can do it, then you can +2 more easily ;)
10:45:52 slaweq Shrews: no problem :) I was one of those who approved broken patch ;)
10:45:59 slaweq frickler: sure, good idea
11:56:27 openstackgerrit Artem Goncharov proposed openstack/openstacksdk master: handle "paginated" argument in test_list properly https://review.openstack.org/638401
12:01:05 slaweq frickler: looks like my patch works and sdk job is happy http://logs.openstack.org/85/638385/1/check/openstacksdk-functional-devstack-networking/a2a55f6/testr_results.html.gz :)
12:08:17 frickler slaweq: cool, good work :-)
12:24:56 slaweq frickler: thx :)
14:59:16 mordred slaweq, frickler: yay!
15:18:00 elmiko edleafe: just a heads up, i have a conflicting meeting today so i will be around but on a high latency delay
15:18:32 elmiko dtantsur too ^^
15:18:34 edleafe elmiko: I'm around, but I have a PT sessions scheduled that will keep me from my computer
15:18:52 elmiko k, i'll try to keep an eye out
15:22:04 dtantsur I also have a conflicting meeting btw
15:22:24 elmiko haha
15:23:03 elmiko ok, i'll keep an eye on the chat
15:56:01 elmiko API SIG office hour starts in 5 minutes, prepare your inquiries!
16:17:04 edleafe ...and I'm back
16:28:59 openstackgerrit Rodolfo Alonso Hernandez proposed openstack/openstacksdk master: Add "name" filter in "list" call when retrieving a single register https://review.openstack.org/637238
16:39:24 cdent elmiko, edleafe : If I have a resource which returns a json object which contains nested json objects, if I have a GETtable subresoruce for one of those nested json objects am I obliged to have subresources for all the nested objects?
16:39:57 elmiko cdent: good question
16:40:03 edleafe "Obliged"?
16:40:28 edleafe It would seem likely that you would need that, but I don't think that you are under any sort of obligation
16:40:32 mordred cdent: when you say nested json objects ...
16:40:39 elmiko yeah "obliged" sounds kinda strong to me, especially if there is no value in exposing the subresources by themselves
16:40:54 mordred (yeah, I agree with edleafe and elmiko on the obligation part)
16:42:18 cdent As I hope is fairly obvious, I don't have a real use case here, I just want office hours to have _something_
16:42:51 dtantsur :)
16:43:04 edleafe Ah, I thought someone had posed such a question to you and you were passing it along
16:43:10 dtantsur I don't think we should create subresources for the sake of having them (we're quite guilty of that in ironic)
16:43:13 elmiko haha, cdent ++
16:43:29 cdent beer?
16:43:41 elmiko i say no to "obliged", yes to "if there is value in exposing the subresource"
16:45:21 mordred cdent: maybe?
16:45:59 cdent I also say no to "obliged" but I know there are some who would feel that way for sake of grammatical parallelism
16:46:28 elmiko i mean, if you have the cycles to burn....
16:52:38 dtantsur everyone here has a lot of spare time, why even ask?
16:52:45 dtantsur :)
16:56:54 elmiko now if you are using automagical tooling to create your server... XD
16:57:42 dtantsur sounds ironic!
16:57:47 elmiko haha
17:12:14 elmiko gonna wander off for some lunch, good chatting with y'all as usual =)
17:13:06 dtantsur enjoy
19:34:51 openstackgerrit David Rabel proposed openstack/python-openstackclient master: Only allow project ID in 'image add project' https://review.openstack.org/638495
19:35:13 openstackgerrit David Rabel proposed openstack/python-openstackclient master: Fix help message of image add project https://review.openstack.org/638113
20:50:21 openstackgerrit Merged openstack/openstacksdk master: Add image attributes from v2.7 https://review.openstack.org/637538
20:53:38 elmiko anyone around who understand pbr reasonably well?
21:05:04 mordred elmiko: yup.
21:13:18 elmiko mordred: so, i'm using it for a new project and trying to work my through the devel cycle kinks. and i've run into this issue where i create a tag, push the dists to pypi, then i'll do work and when i do a `pip install -U .` of my devel stuff pbr complains about semvers and git stuff. have you run into this?
21:15:12 mordred elmiko: I have not run in to that although that makes me have sads :(
21:15:52 elmiko mordred: definitely giving me a case of the sads
21:16:02 elmiko i can work around it, just curious why it happens
21:17:14 mordred elmiko: I'm trying to think of a good explanation of why that would happen - if you have the git repo locally, that should Just Work (although fwiw I tend to just do a pip install -e . in my source dir and then months later forget that I'm running off of my local branch :) )
21:25:32 elmiko mordred: ack, i am working locally and usually this works. i think i must have futzed with the versions or something.
21:26:14 elmiko what does the `-e` flag do?
21:27:46 mordred elmiko: it does an 'editable' install - it's like python setup.py develop
21:28:00 elmiko mordred: interesting, thanks!
21:28:03 mordred elmiko: so, basically, it installs links to your source dir - and as you change python files it's immediately reflected
21:28:13 elmiko that is super useful
21:28:29 mordred if you add new entrypoints in setup.cfg you'll need to re-run pip install -e .
21:28:37 mordred just fwiw
21:33:38 elmiko ack, thanks again =)
21:33:43 elmiko mordred ++
21:44:29 njohnston Hi! I have a quick question about adding functionality into the openstack sdk. I want to create functionality in the sdk that will use the neutron bulk port creation method, so that you can supply an array of port information in a single request and all of those ports will be created at once and in streamlined manner. Will I need to use something other than the generic _create, or can it handle a list of inputs in and a
21:44:29 njohnston list of objects returned?
21:49:18 mordred njohnston: that is a a great question! I don't think the generic _create will do what you want - so you might need to just do your own magic method
21:49:47 mordred njohnston: that said - maybe once you've got that we can abstract out a _create_list or something like that
21:50:17 njohnston mordred: Great, that is what I thought! I'll start working on it. Thanks!
21:51:40 mordred \o/
22:11:45 njohnston mordred: If I am understanding this code right, it sounds like I might need a different fundamental class than Resource, something like BulkResource that understands that instead of getting **attrs and returning Resource will be getting [attrs, attrs, attrs] and returning [Resource, Resource, Resource]. Does that sound like a sane proposition?
22:18:33 mordred njohnston: maybe? but also - maybe it's just a method on the proxy? (like, making a BulkResource might be a significant engineering effort - and the main bits of Resource are largely about expressing the remote objects
22:19:05 mordred njohnston: that said - maybe you do want a BulkResource object that takes another resource class as a parameter and does like you said
22:20:16 mordred njohnston: so that instead of conn.network.create_ports([args, args, args]) returning [Port, Port, Port] it returns BulkResource which wraps a list of Ports?
22:20:43 mordred njohnston: I could go either way - might be one of those "get in to it and see which sucks less" sorts of things
22:24:48 njohnston yeah, sounds like it. :-)

Earlier   Later