Earlier  
Later
New Search
Posted Nick Remark
#openstack-sdks - 2019-05-23
17:00:58 elmiko take care edleafe, have a nice holiday weekend =)
17:03:39 edleafe I shall! You too!
17:44:30 openstackgerrit Logan V proposed openstack/openstacksdk master: Update Limestone Networks vendor config https://review.opendev.org/661063
17:45:00 openstackgerrit Logan V proposed openstack/openstacksdk master: Update Limestone Networks vendor config https://review.opendev.org/661063
17:45:51 openstackgerrit Logan V proposed openstack/openstacksdk master: Update Limestone Networks vendor config https://review.opendev.org/661063
20:34:53 cmurphy edleafe: elmiko cdent my issue is that if more than one error is returned, eg if nova fails because glance fails, I don't think it's the user's business to know what happened on the backend, it could even leak sensitive information, so it should just be nova's job to capture the error and reformat it as one error for the user, the operator has logs to see what really happened on the backend
20:49:13 mordred cmurphy, edleafe, elmiko: it can be confusing as well - because those errors are not actionable. I know we've gotten that at times in Infra - an error booting a node because there was some issue between nova and glance that had nothing to do with our boot command or the image contents
20:49:51 mordred so it tells us "blah blah cna't find /dev/xx05a blah blah ceph blah" - and it's both scary and makes us try to troubleshoot why we were trying to do something with ceph - and it turns out we're not we're just trying to boot a vm
20:50:03 edleafe cmurphy: I don't think that you should ever return internal details in a user-facing API, whether it's a single message or multiple. Your example of glance-level errors should only be on admin APIs
20:50:29 mordred edleafe: should and are have so many differences
20:50:41 mordred edleafe: (but I agree with you)
20:50:45 edleafe mordred: Guidelines are for "should"
20:50:49 edleafe :)
20:51:06 mordred edleafe: life is friendly in should
20:52:30 edleafe cmurphy: But to get back to your point: error messages and details should be informative. Packing every known bit of data in an error messages is just as bad as not including enough
20:53:04 edleafe cmurphy: So if there are more than one relevant bits of information about the error, they should be included, not just one of them
20:53:44 edleafe The guideline is more about how to handle multiple bits of information. It doesn't require that you include more than one, though
21:41:03 elmiko cmurphy: i don't think the intent was for a single service to bubble out errors from other services. i had always thought of it as a single service emitting multiple errors to a single call.
21:42:31 elmiko and i agree with edleafe re: the api not /requiring/ more than one
23:30:55 openstackgerrit Matt Riedemann proposed openstack/python-openstackclient master: WIP: Deprecate openstack server migrate --host option https://review.opendev.org/659382
#openstack-sdks - 2019-05-24
00:11:03 openstackgerrit Matt Riedemann proposed openstack/python-openstackclient master: WIP: Deprecate openstack server migrate --host option https://review.opendev.org/659382
06:15:02 cmurphy edleafe: elmiko do you have an example of an API call that would emit more than one user-facing error for a single request?
06:15:27 mfeder dtroyer/mordred: Hi guys, Is there any chance to move forward with this one https://review.opendev.org/#/c/576507/ ? Thanks for answer.
09:35:41 openstackgerrit Merged openstack/openstacksdk master: Update Limestone Networks vendor config https://review.opendev.org/661063
09:55:04 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Blacklist bandit 1.6.0 & cap sphinx for 2.7 https://review.opendev.org/660956
10:20:59 openstackgerrit Vishakha Agarwal proposed openstack/keystoneauth master: Blacklist bandit 1.6.0 & cap sphinx for 2.7 https://review.opendev.org/660956
13:14:20 openstackgerrit Dean Troyer proposed openstack/python-openstackclient master: Add changes-before attribute to server list https://review.opendev.org/657373
14:47:39 Shrews mordred: did you get any info on that os_port PR?
14:48:08 Shrews weird that it hasn't merged
15:00:11 openstackgerrit Matt Riedemann proposed openstack/python-openstackclient master: Deprecate openstack server migrate --host option https://review.opendev.org/659382
15:02:17 openstackgerrit Matt Riedemann proposed openstack/python-openstackclient master: Deprecate openstack server migrate --host option https://review.opendev.org/659382
15:26:02 mordred Shrews: yes! see #ansible-devel - also useful search link there
15:28:24 Shrews mordred: i do not understand how the CL fragment prevented merge
15:28:54 mordred Shrews: I think automerge is disabled if files outside of the modules dir were affected
15:29:33 Shrews is it bad i hate that entire CI system?
19:58:03 openstackgerrit Matt Riedemann proposed openstack/python-openstackclient master: Deprecate openstack server migrate --host option https://review.opendev.org/659382
21:13:49 edleafe cmurphy: Sorry, just saw your question from yesterday. No, I don't have an example off-hand, but the AWS docs on errors seem to imply that they support multiple errors. See the very bottom of https://docs.aws.amazon.com/AWSEC2/latest/APIReference/errors-overview.html
21:14:08 edleafe Actually, that's https://docs.aws.amazon.com/AWSEC2/latest/APIReference/errors-overview.html#api-error-response
#openstack-sdks - 2019-05-27
06:30:08 openstackgerrit Brin Zhang proposed openstack/python-openstackclient master: Add changes-before attribute to server list https://review.opendev.org/657373
12:05:39 openstackgerrit Brin Zhang proposed openstack/python-openstackclient master: openstack port create support --extra-dhcp-option https://review.opendev.org/657519
13:52:50 ITD27M01 mordred: Hello! If you have a minute could you please participate in discussion and review in https://github.com/ansible/ansible/pull/51401
#openstack-sdks - 2019-05-28
02:58:08 openstackgerrit ZhouHeng proposed openstack/python-openstackclient master: openstack port create support --extra-dhcp-option https://review.opendev.org/657519
03:03:16 openstackgerrit ZhouHeng proposed openstack/python-openstackclient master: openstack port create support --extra-dhcp-option https://review.opendev.org/657519
07:08:39 openstackgerrit Brin Zhang proposed openstack/python-openstackclient master: Add changes-before attribute to server list https://review.opendev.org/657373
14:21:37 openstackgerrit Brian Haley proposed openstack/python-openstackclient master: openstack port create support --extra-dhcp-option https://review.opendev.org/657519
20:24:13 openstackgerrit Merged openstack/keystoneauth master: Blacklist bandit 1.6.0 & cap sphinx for 2.7 https://review.opendev.org/660956
23:00:29 openstackgerrit Clark Boylan proposed openstack/python-openstackclient master: Replace pkg_resources with importlib-metadata https://review.opendev.org/661856
23:09:44 openstackgerrit Clark Boylan proposed openstack/python-openstackclient master: Replace pkg_resources with importlib-metadata https://review.opendev.org/661856
#openstack-sdks - 2019-05-29
02:30:15 openstackgerrit jacky06 proposed openstack/keystoneauth master: Replace git.openstack.org URLs with opendev.org URLs https://review.opendev.org/655016
08:39:51 openstackgerrit Andreas Jaeger proposed openstack/openstackclient master: Remove pbrx https://review.opendev.org/661914
11:55:19 openstackgerrit Akihiro Motoki proposed openstack/openstackclient master: Update sphinx dependency for python 2.7 https://review.opendev.org/661947
12:40:24 AJaeger mordred: for fixing openstackclient, you need to fix sphinx first - see https://review.opendev.org/#/c/661947/ and then the py27 bug in there ;(
12:56:14 AJaeger amotoki, mordred, I pushed a change on top of 661947. Interesting, you do not publish docs - is that intentional?
13:02:55 openstackgerrit Andreas Jaeger proposed openstack/openstackclient master: Update sphinx dependency for python 2.7 https://review.opendev.org/661947
13:02:55 openstackgerrit Andreas Jaeger proposed openstack/openstackclient master: Switch to openstackdocstheme https://review.opendev.org/661974
13:03:06 AJaeger amotoki: updated your change
13:08:10 elmiko cmurphy: in answer to your last question, when i went to look back through the comments on the initial guideline proposal, it /looked/ lke someone had posted an example of returning multiple errors to a single call. outside of that though, i don't have an example at the ready.
13:23:43 openstackgerrit Andreas Jaeger proposed openstack/openstackclient master: Update sphinx dependency for python 2.7 https://review.opendev.org/661947
13:23:44 openstackgerrit Andreas Jaeger proposed openstack/openstackclient master: Switch to openstackdocstheme https://review.opendev.org/661974
13:57:05 AJaeger mordred, dtroyer, so https://review.opendev.org/661947 fixes all the failures that openstackclient currently has and is the next change you need to merge...
13:59:17 amotoki AJaeger: thanks for the update. As you pointed, we need to check the necessity of doc dependency and I will check it, but it looks good to merge 661947 first as it affects various changes.
14:00:42 AJaeger amotoki: yes, we can discuss 661974 separately. Please discuss whether you want to publish docs, only build them - or just take the change without jobs to install less packages ;)
14:50:30 cmurphy elmiko: what i still find strange is that this would all be stuffed into the body of a single response, while the actual http response would only have one error code
14:51:13 cmurphy it looks like rfc 7230 says that multiple responses could be returned for a single request but that's not how i read the guideline
15:23:15 elmiko cmurphy: i think you are right that it does seem odd for the http request to return a single error, but the body could express multiple internal errors
15:23:43 elmiko and i would imagine in most cases the error in the body would reflect a single error about the return code
15:24:15 elmiko if memory serves me correctly, and it might not, i think the idea was that a 500 error could expose more internal detail through this mechanism
15:24:28 elmiko obviously for something like 404 it doesn't make much sense
15:25:14 elmiko and yeah, i think the intent in 7230 is different than what we are describing in the guideline
16:01:06 cdent cmurphy, elmiko: the goal, as I remember, is to encapsulate the fact that there could be a suite of several system errors that led up to a particular http error, and knowing both is desirable, thus the sepraration between and status and code
16:01:14 cdent and yeah 7230 is something else
16:01:33 cdent not that I'm trying to defend the guideline. I've always thought it a bit weird, just trying to dredge up the reasons
16:02:39 openstackgerrit Clark Boylan proposed openstack/python-openstackclient master: Replace pkg_resources with importlib-metadata https://review.opendev.org/661856
16:09:47 cmurphy as i said in the thread i don't mind patching keystoneauth to conform to the guideline, i just think the guideline is weird
16:18:26 elmiko cmurphy cdent, for clarity sake, i'm not trying to defend the guideline either, and it may in fact be weird.
16:18:51 elmiko this does provide an excellent opportunity to reflect on how we might change things when turn out to not make sense
16:19:23 elmiko i'm curious if we need to re-open that guideline and discuss it again if it turns out that a majority think it's weird and will probably never implement it
16:20:06 cdent i suspect, unfortunately, that now that implementations of multiple styles are lose in the wild, we need to support them both forever
16:20:33 cdent because though we've been willing to say that message in errors are okay to change whenever, changing their structure probably less so
16:59:00 openstackgerrit Merged openstack/keystoneauth master: Resolves a typo in a link to use Application Credentials https://review.opendev.org/660886
18:45:09 openstackgerrit Merged openstack/keystoneauth master: Replace git.openstack.org URLs with opendev.org URLs https://review.opendev.org/655016
18:50:20 mordred if cdent was here, I'd say "of course, there are multiple forms of error payloads, so some things, like openstacksdk for instance, just do a shotgun approach and try their best to find some structure that might have error information in it"
#openstack-sdks - 2019-05-30
14:01:31 ITD27M01 mordred: I really need your attention. If you have a minute, look at the https://github.com/ansible/ansible/pull/51401
16:00:03 elmiko API SIG office hour is now open! \o/
16:00:03 edleafe Welcome to the API-SIG Office Hour!
16:00:08 elmiko jinx!
16:00:11 edleafe Heh, you beat me!
16:00:25 elmiko it was like simultaneous on my end lol
16:01:34 edleafe Mine got recorded first here: https://leafe.com/timeline-middle/%23openstack-sdks/2019-05-30T16:00:03
16:01:41 edleafe So we both win!
16:01:52 elmiko hahaha, amazing
16:15:23 cdent I lost track, did the errors guidelines discussion go anywhere?
16:15:56 edleafe There was a little async back-and-forth in this channel, but it didn't really go anywhere
16:16:09 elmiko ++
16:16:12 cmurphy o/
16:16:39 cmurphy i don't think there's much more to discuss, it is what it is
16:16:56 edleafe cdent: Starting here: https://leafe.com/timeline-middle/%23openstack-sdks/2019-05-23T20:34:53
16:18:53 cdent is there a next step?
16:19:17 elmiko unless we are going to rewrite the guideline, i don't think so
16:19:32 cdent potentially a patch to keystoneauth?

Earlier   Later