| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-sdks - 2020-05-06 | |||
| 19:30:33 | dtroyer | except I am likely to resurrect the devstack technique I tried for running osc in the background and piping comamnds to it… we have a rather annoying auth situation here, which is why cutting the /auth/token round trips in half is worthwhile | |
| 19:31:07 | dtroyer | and for my next trick, learning to do the same for fog… | |
| 19:32:01 | mordred | dtroyer: I don't suppose getting the cloud to update and add appcred support is likely then | |
| 19:32:41 | dtroyer | It's lucky to be on queens (vmware-backed) and they don't even know what that is. | |
| 19:33:59 | mordred | dtroyer: well - I guess our insistence on supporting old openstacks has paid off | |
| 19:35:05 | dtroyer | absofrickinlutely…. never give up that fight :) | |
| 19:36:05 | dtroyer | mordred: while I've got you, what did you guys switch to rather than cloud-init? | |
| 19:36:21 | mordred | we use glean | |
| 19:36:30 | mordred | https://opendev.org/opendev/glean | |
| 19:36:44 | mordred | if you build images with dib, there is a "simple-init" element that sets it up | |
| 19:37:20 | mordred | https://opendev.org/openstack/diskimage-builder/src/branch/master/diskimage_builder/elements/simple-init | |
| 19:37:24 | dtroyer | ah, right. thanks. this is an existing chef/packer/rhel stack that can't configure networking properly without help | |
| 19:37:41 | mordred | nod. you should be able to install it with packer | |
| 19:37:53 | mordred | (but you might want to go crib systemd units from that element) | |
| 19:38:12 | mordred | it's limited in scope - it can do networking and ssh keys and only reads data from config-drive | |
| 19:38:51 | mordred | but - it can handle static network config in config-drive | |
| 19:39:30 | dtroyer | apparently cloud-init on rhel doesn't do that well at all | |
| 19:48:26 | mordred | yeah - static networking from cloud-init is why we wrote glean | |
| 19:49:29 | mordred | rax only does static networking (no dhcp) and only does config-drive - and for quite a long time cloud-init dev was frozen because they were working on a v2 so we coudlnt' get the support for it added | |
| 19:49:29 | dtroyer | and I love that dependency list! | |
| 19:49:37 | mordred | (there was a rax-specific fork of cloud-init - but we didn't really feel like that was the right life decision) | |
| 19:50:26 | mordred | and yeah - also - cloud-init pulls in a bunch of python, and we happen to, you know, be in the business of testing a lot of python | |
| #openstack-sdks - 2020-05-07 | |||
| 07:27:00 | openstackgerrit | Johannes Kulik proposed openstack/openstacksdk master: Add "id" to Port's query parameters https://review.opendev.org/726050 | |
| 07:29:11 | openstackgerrit | Johannes Kulik proposed openstack/openstacksdk master: Add "id" to Port's query parameters https://review.opendev.org/726050 | |
| 09:11:17 | openstackgerrit | Riccardo Pittau proposed openstack/osc-lib master: Remove six library from dependencies https://review.opendev.org/724968 | |
| 13:10:58 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Use floating ip for interface ip if it exists https://review.opendev.org/726115 | |
| 13:11:30 | openstackgerrit | Monty Taylor proposed openstack/openstacksdk master: Use floating ip for interface ip if it exists https://review.opendev.org/726115 | |
| 13:41:06 | nightmare_unreal | how does OSC CLI determine which api version to use ? | |
| 13:41:28 | nightmare_unreal | for e.g. use 2.15 or something or 2.60 etc etc. | |
| 14:17:02 | openstackgerrit | Merged openstack/shade stable/ussuri: Break out py27 job templates https://review.opendev.org/723353 | |
| 14:23:02 | openstackgerrit | Merged openstack/shade stable/ussuri: Update .gitreview for stable/ussuri https://review.opendev.org/719019 | |
| 14:23:03 | openstackgerrit | Merged openstack/shade stable/ussuri: Update TOX/UPPER_CONSTRAINTS_FILE for stable/ussuri https://review.opendev.org/719020 | |
| 14:40:06 | openstackgerrit | Merged openstack/osc-lib master: Remove six library from dependencies https://review.opendev.org/724968 | |
| 14:47:03 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Fix metric names in the object_store https://review.opendev.org/726167 | |
| 14:54:55 | openstackgerrit | Artem Goncharov proposed openstack/openstacksdk master: Fix metric names in the object_store https://review.opendev.org/726167 | |
| 16:01:12 | elmiko | API SIG office hour open | |
| 16:10:46 | brtknr | hmm why doesnt pip2 understand not to install openstacksdk version greater than 0.40.0 by default | |
| 16:40:44 | mordred | brtknr: 0.45 is the last py2 version | |
| 16:42:02 | mordred | brtknr: we dropped 2.7 from the gate in 0.40 but didn't get the python-requires line in. we accidentally broke 2.7 in 0.44, reverted the patch and released 0.45, then added the metadata in 0.45 to tell pip2 to do the right thing | |
| 16:42:37 | mordred | brtknr: are you have having issues with 0.45 and python 2.7? if so, we should figure them out and see if we can't figure out how to release an 0.45.1 point release | |
| 16:42:49 | mordred | but otherwise, you should get 0.45 and it should work | |
| 16:59:23 | elmiko | catch y'all later, stay safe out there o/ | |
| 17:43:35 | dtroyer | Are there any known issues with uploading images to vmware-backed clouds since the image code was switched to SDK? I am getting persistent checksum failures with osc 5+ but not 4.0.0 | |
| 18:40:02 | mordred | dtroyer: no, not to my knowledge. that said - I do wanna say we found $something related to checksums uploading to $somewhere - but I thought it was fixed | |
| 18:42:38 | mordred | I'm not finding anything ... | |
| 18:43:21 | mordred | gtema: ^^ | |
| 18:43:30 | mordred | gtema: do you remember something with that? | |
| 18:47:11 | dtroyer | mordred: ok, thanks. I'll put together a repeatable command and get some logs. The same image uploaded to DreamCompute fine with basically the same command and osc venv. | |
| 18:48:21 | mordred | dtroyer: *awesome* | |
| 18:49:03 | mordred | dtroyer: maybe there's a transform happening cloud-side or something | |
| 18:50:40 | dtroyer | I do know it is VMware's Queens-era vSphere+OpenStack whatever-they-call-it but that's about it | |
| 18:50:52 | mordred | dtroyer: clarkb says: | |
| 18:50:54 | mordred | clarkb>mordred: ish. They dont do anything useful except for the swift based upload dedup | |
| 18:51:03 | mordred | clarkb>mordred: glance can change the image without telling you and so thr checksum you provide to glance is uselss also glance doesnt check it aiui | |
| 18:51:33 | mordred | dtroyer: so it's possible we're providing a checksum to glance now which is potentially problematic because $sadness | |
| 18:52:18 | dtroyer | I suppose it should be noted too that these are vmdk images, which I've never uploaded to Glance before, dunno if there is anything to that | |
| 18:52:36 | mordred | dtroyer: if you find that to be true, I'd definitley be in support of just not providing it if it's garbage with no value and only serves to possible break people | |
| 18:53:45 | mordred | dtroyer: look for validate_checksum in openstack/image/v2/_proxy.py | |
| 18:53:51 | dtroyer | when I loop back around to this I can tweak that out to see if it matters…kk | |
| 18:53:51 | dtroyer | when I loop back around to this I can tweak that out to see if it matters…kk | |
| 18:54:18 | mordred | dtroyer: I think as an easy first step - we can probably just set that to false in osc | |
| 18:54:35 | mordred | dtroyer: and if that fixes it - maybe we change the default value in sdk to false | |
| 18:55:23 | mordred | dtroyer: when the cloud backend is swift and we have to upload to swift first, we use the locally calculated checksums as a de-duplication step so that if yo have to upload again you can avoid re-uploading swift objects | |
| 18:55:34 | mordred | but I thnik for non-swift it really is just a waste of energy | |
| 18:55:58 | mordred | and even in the swift case I don't think the glance checksum helps anything | |
| 19:52:40 | openstackgerrit | Merged openstack/openstacksdk master: Add cipher list support for octavia https://review.opendev.org/721351 | |
| 21:38:55 | brtknr | mordred: weirdly that issue has magically disappeared now. not sure why it was there in the first place | |
| 21:39:24 | brtknr | its working for me now, nothing to worry about | |
| 21:59:02 | mordred | brtknr: \o/ | |
| #openstack-sdks - 2020-05-08 | |||
| 10:04:43 | openstackgerrit | Merged openstack/openstacksdk master: Fix metric names in the object_store https://review.opendev.org/726167 | |
| 14:47:02 | nightmare_unreal | how does osc cli determine which api version to use ?? | |
| 15:30:50 | openstackgerrit | Alex Schultz proposed openstack/openstacksdk master: Don't error if clouds.yaml is not readable https://review.opendev.org/726416 | |
| 15:44:54 | openstackgerrit | Sagi Shnaidman proposed openstack/ansible-collections-openstack master: DNM test q branch https://review.opendev.org/726421 | |
| 16:03:23 | gtema | mordred, we don't have a PTG slot planned for SDK or API-SIG? Or am I blind | |
| 16:06:29 | gtema | dtantsur, elmiko ^^. Deadline seems to be this Sunday | |
| 16:07:22 | elmiko | we put forth no planned session for the api-sig | |
| 16:07:45 | elmiko | is there a desire to have one? (i feel like we are running on fumes at this point, activity-wise) | |
| 16:08:03 | gtema | we were talking to have some | |
| 16:08:13 | gtema | discovery documents, etc | |
| 16:08:45 | elmiko | gtema: if you want to take the lead on proposing a session for api-sig, i'm +2 for it =) | |
| 16:08:46 | gtema | and definitely on SDK part | |
| 16:08:58 | elmiko | i figured sdk would have something to talk about | |
| 16:10:36 | gtema | okay, filling the form | |
| 16:15:04 | gtema | does anyone knows, whether it's only about filling form (we want), or also the time planning in ethercalc? | |
| 16:17:49 | elmiko | i'm not sure | |
| 16:18:13 | gtema | I'm also lost in all those emails | |
| 16:18:41 | elmiko | maybe reach out to diablo_rojo_phon ? | |
| 16:18:48 | smcginnis | I think the ethercalc is the important step. | |
| 16:19:15 | gtema | okay, will do so | |
| 16:19:47 | gtema | I guess for both API-SIG and SDK/CLI we are mostly Europe+US, right? | |
| 16:22:37 | gtema | those ironic guys, blocked almost each day ;-) | |
| 16:27:45 | gtema | entered 1 hour for API-SIG June 2nd 16 - 17 UTC, and for SDK/CLI: June 1st 15-17 UTC and 3rd 15-17 UTC | |
| 16:32:06 | diablo_rojo_phon | gtema: make sure to fill out the survey too and not just the ethercalc :) | |
| 16:32:16 | gtema | done already, thans | |
| 16:32:18 | gtema | thanks | |
| 16:43:26 | diablo_rojo_phon | Thank you! | |
| 16:45:13 | elmiko | gtema: thanks for taking the lead on that =) | |
| 17:56:35 | dtroyer | mordred, gtema: Setting validate_checksum=False fixed the image create issue with the vmware cloud… It is defaulted True in 3 places in the SDK, changing only BaseImageProxy.create_image is sufficient, or of course turning it off in OSC would work too. Do you guys have a preference? | |
| 17:56:35 | dtroyer | mordred, gtema: Setting validate_checksum=False fixed the image create issue with the vmware cloud… It is defaulted True in 3 places in the SDK, changing only BaseImageProxy.create_image is sufficient, or of course turning it off in OSC would work too. Do you guys have a preference? | |
| 18:08:55 | mordred | dtroyer: I think changing it in BaseImageProxy.create_image should be fine, yeah? I think we should just do that - best I can tell it's not super useful | |