openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 00:04 |
---|---|---|
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 01:41 |
*** Sukhdev has joined #openstack-dib | 02:34 | |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 03:26 |
*** Sukhdev has quit IRC | 03:33 | |
*** pbourke_ has quit IRC | 04:05 | |
*** pbourke_ has joined #openstack-dib | 04:06 | |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 04:24 |
*** eshas has joined #openstack-dib | 04:52 | |
*** eshas has quit IRC | 05:38 | |
*** eshas has joined #openstack-dib | 05:39 | |
*** andreas_s has joined #openstack-dib | 06:41 | |
*** hashar has joined #openstack-dib | 06:50 | |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 07:00 |
olivierb- | hello everybody, are aome/all images produced by CI tests in DIB checked for properly booting/running ? I am more specifically wanted to see some LVM based ones | 07:07 |
*** brault has quit IRC | 07:07 | |
ianw | oliverb-: a large subset, but not with lvm | 07:39 |
ianw | these are the "nodepool" tests in the CI results. we use devstack to bring up an openstack environment, then nodepool to build, upload & boot images | 07:39 |
ianw | we do everything that openstack uses (ubuntu's, centos, fedora, suse's, gentoo) | 07:40 |
openstackgerrit | Ian Wienand proposed openstack/diskimage-builder master: Remove installed packages before pip install https://review.openstack.org/561479 | 07:44 |
olivierb- | ianw thx for this, my current experience with LVM is somewhat controversial right now due to some very weird behaviour therefore my question | 07:46 |
ianw | oliverb-: yeah, it has honestly only been lightly used, i would not be surprised if you find things we don't handle | 07:47 |
*** eshas has quit IRC | 07:48 | |
*** pavel_V has joined #openstack-dib | 08:02 | |
*** brault has joined #openstack-dib | 08:36 | |
*** eshas has joined #openstack-dib | 08:37 | |
*** brault_ has joined #openstack-dib | 09:00 | |
*** brault has quit IRC | 09:02 | |
*** eshas has quit IRC | 09:54 | |
*** eshas has joined #openstack-dib | 09:54 | |
pbourke_ | ianw: hi Ian, I seem to be getting the following after updating dib this morning: | 10:03 |
pbourke_ | "Cannot uninstall 'virtualenv'. It is a distutils installed project and thus we cannot accurately determine which files belong to it which would lead to only a partial uninstall." | 10:03 |
pbourke_ | ianw: anything to do with Ia84c38dc3c40a6080e144b563e10abca7dac2881 by any chance? | 10:04 |
pbourke_ | command im using: disk-image-create --ramdisk-element=dracut-ramdisk --logfile out -t tgz -o deploy.ramdisk ironic-agent centos7 | 10:04 |
*** pbourke_ is now known as pbourke | 10:05 | |
ianw | pbourke: likely https://review.openstack.org/561479 will fix it | 10:08 |
pbourke | ianw: thanks will check it out | 10:09 |
*** esha1 has joined #openstack-dib | 11:35 | |
*** eshas has quit IRC | 11:35 | |
*** esha1 has quit IRC | 12:08 | |
*** vmlinuz has joined #openstack-dib | 12:42 | |
*** vmlinuz has joined #openstack-dib | 12:42 | |
*** eshas has joined #openstack-dib | 12:42 | |
*** mjturek has joined #openstack-dib | 12:54 | |
openstackgerrit | Bharat Kunwar proposed openstack/diskimage-builder master: Fix yaml dependency issue with dracut-regenerate for python3 https://review.openstack.org/558497 | 13:30 |
*** eshas has quit IRC | 14:03 | |
*** eshas has joined #openstack-dib | 14:38 | |
openstackgerrit | Bharat Kunwar proposed openstack/diskimage-builder master: Fix yaml dependency issue with dracut-regenerate for python3 https://review.openstack.org/558497 | 14:49 |
*** rajinir has joined #openstack-dib | 14:59 | |
*** vmlinuz has quit IRC | 15:10 | |
*** vmlinuz has joined #openstack-dib | 15:10 | |
*** vmlinuz has joined #openstack-dib | 15:10 | |
*** hashar is now known as hasharAway | 15:16 | |
*** andreas_s has quit IRC | 15:42 | |
*** andreas_s has joined #openstack-dib | 15:43 | |
eshas | hi, The ubuntu cloud image .img can be directly downloaded and converted to RAW by qemu command and then booted | 15:49 |
eshas | why do we need DIB then? | 15:49 |
prometheanfire | eshas: troll? | 15:50 |
*** mgagne_ is now known as mgagne | 15:50 | |
eshas | huh? | 15:50 |
prometheanfire | eshas: dib doesn't do just simple images, it is a build system | 15:51 |
prometheanfire | doesn't do just ubuntu either | 15:51 |
clarkb | right, the idea is it allows you to customize existing images in a reproduceable and fast manner that doesn't require virt capabilities. It also allows you to build new images from scratch without an existing image | 15:52 |
eshas | yes, but for rhel and ubuntu as cloud images are available, and they have OS ready, then if I download them in a ppc64le env and qemu create to RAW and chroot and boot them | 15:52 |
prometheanfire | eshas: and they don't provide things like glean, pip, caching openstack repos, etc | 15:53 |
eshas | then am good to go right..unless I need cloud init and other customization | 15:53 |
eshas | clarkb: how does it build new images from scratch? | 15:54 |
prometheanfire | eshas: https://docs.openstack.org/diskimage-builder/latest/ | 15:55 |
eshas | <prometheanfire>: ok, probably need to see how to do customization using DIB for cloud-init , epel, ssh,networking etc then it might be useful | 15:55 |
eshas | hmm..I know its working but thought it uses cloud images always | 15:56 |
eshas | for ubuntu and RHEL | 15:56 |
prometheanfire | nope | 15:56 |
eshas | as iso not possible | 15:56 |
eshas | to pass to it | 15:56 |
*** andreas_s has quit IRC | 15:57 | |
*** andreas_s has joined #openstack-dib | 16:02 | |
clarkb | eshas: for ubuntu and debian it runs debootstrap directly and builds the image from nothing | 16:04 |
eshas | what is meant by nothing? it takes the cloud img on disk-image-create | 16:05 |
eshas | downloads it as per release | 16:05 |
prometheanfire | nope | 16:06 |
prometheanfire | https://github.com/openstack/diskimage-builder/blob/master/diskimage_builder/elements/debian-minimal/root.d/75-debian-minimal-baseinstall | 16:06 |
prometheanfire | https://github.com/openstack/diskimage-builder/blob/master/diskimage_builder/elements/debootstrap/root.d/08-debootstrap#L33 | 16:07 |
prometheanfire | not a cloud image | 16:07 |
prometheanfire | please read the docs/code | 16:07 |
clarkb | eshas: prometheanfire it depends on what element you choose. One option is to start from an existing image. Another is to build it from nothing | 16:08 |
clarkb | eshas: starting from an existing image is simple because someone else has done a lot of the work for you. But the control of being able to build exactly what you want can be useful in some cases | 16:08 |
eshas | yes, for that kind of complete control how do i start? | 16:09 |
clarkb | eshas: use the -minimal elements | 16:09 |
eshas | cause currently we use iso and then install OS and customise cloud init etc | 16:09 |
prometheanfire | I linked the docs earlier | 16:09 |
eshas | ok ok | 16:10 |
eshas | I am running into ssh / login issue with ubuntu via DIB | 16:10 |
eshas | :( | 16:10 |
prometheanfire | look at the devuser element for debug users | 16:11 |
eshas | so was thinking if mounting and chroot will help | 16:11 |
eshas | I did create ubuntu 18.04 raw via DIB, didnt use devuser though | 16:11 |
eshas | was trying ssh keys but didnt work | 16:11 |
eshas | will try again with devuser defined | 16:12 |
eshas | Another thing I was trying was to run DIB in a non KVM environment but openstack based | 16:12 |
eshas | I installed debootstrap separately as DIB reported it missing | 16:12 |
eshas | but still more issues seen | 16:13 |
eshas | need to use DIB in this ppc64le non-kvm openstack env | 16:13 |
clarkb | eshas: yes the docs should tell you what the system reqs are for building those images | 16:13 |
clarkb | you will need debootstrap and apt and so on | 16:13 |
openstackgerrit | Félix Bouliane proposed openstack/diskimage-builder master: Allow to pin ubuntu version rather than current https://review.openstack.org/559167 | 16:16 |
*** andreas_s has quit IRC | 16:20 | |
*** andreas_s has joined #openstack-dib | 16:25 | |
*** eshas has quit IRC | 16:33 | |
*** andreas_s has quit IRC | 16:39 | |
*** andreas_s has joined #openstack-dib | 16:44 | |
*** eshas has joined #openstack-dib | 16:48 | |
eshas | clarkb: which docs? | 16:49 |
clarkb | eshas: https://docs.openstack.org/diskimage-builder/latest/user_guide/installation.html has the generic requirements but points to specific elements docs for their system requirements. In this case ubuntu-minimal should say | 16:50 |
eshas | ok, got it, thanks, will check ubuntu and rhel7 i guess | 16:50 |
eshas | as I need for both rhel7.5 and ubuntu18.04 | 16:51 |
clarkb | eshas: well specifically ubuntu-minimal not ubuntu if you want to build from nothing | 16:51 |
eshas | oh, but the images can be used as they are ppc64le compliant I thought, just need to make DIB work in the env | 16:51 |
clarkb | its up to you | 16:52 |
eshas | however shall try step wise then minimal first then full | 16:52 |
eshas | thanks for the tip | 16:52 |
eshas | for rhel there is no such minimal right | 16:52 |
clarkb | eshas: not for rhel but there is for centos | 16:52 |
eshas | hmm.. no I need only rhel sles and ubuntu | 16:53 |
eshas | sles cant be done by DIB | 16:53 |
eshas | so the other two | 16:53 |
clarkb | sles likely can be done by dib you may have to implement the element for it though | 16:53 |
eshas | but sles cloud image for ppc64le doesnt exist | 16:54 |
eshas | atleast I could nt find it | 16:54 |
clarkb | right you don't have to start from an existing cloud image typically | 16:56 |
*** andreas_s has quit IRC | 16:58 | |
eshas | hmm | 16:58 |
eshas | not iso and not cloud image then ? | 16:59 |
clarkb | correct like with ubuntu-minimal you run the distro bootstrap in a chroot and then copy that to an image | 17:03 |
*** calebb has quit IRC | 17:04 | |
*** calebb has joined #openstack-dib | 17:15 | |
*** eshas has quit IRC | 17:28 | |
*** esha1 has joined #openstack-dib | 17:28 | |
openstackgerrit | Félix Bouliane proposed openstack/diskimage-builder master: Allow to pin ubuntu version rather than current https://review.openstack.org/559167 | 17:32 |
*** openstackgerrit has quit IRC | 19:05 | |
*** dhill__ has joined #openstack-dib | 19:22 | |
*** dhill_ has quit IRC | 19:24 | |
*** dhill__ has quit IRC | 19:25 | |
*** dhill_ has joined #openstack-dib | 19:25 | |
*** esha2 has joined #openstack-dib | 19:54 | |
*** esha1 has quit IRC | 19:58 | |
*** hasharAway has quit IRC | 20:01 | |
*** esha2 has quit IRC | 20:24 | |
*** mtreinish has quit IRC | 20:51 | |
*** mtreinish has joined #openstack-dib | 20:57 | |
ianw | clarkb: i know things are busy, wouldn't mind your eye over https://review.openstack.org/#/c/561479/ to use pip10 | 21:11 |
clarkb | ianw: I think that will likely break many jobs for us. Unless we fix tox-siblings first. https://review.openstack.org/#/c/561659/ | 21:12 |
clarkb | mordred hasn't been able to weigh in on ^yet though | 21:12 |
clarkb | I'll review the dib change since I Guess it won't take effect until we release dib | 21:12 |
ianw | clarkb: right, just the presence of pip10? | 21:13 |
clarkb | ianw: ya those jobs work beacuse we are using pip9 in them until our images update successfully with pip10 | 21:13 |
clarkb | ( think pip10 is why our ubuntu images aren't updating/) | 21:14 |
ianw | hmm, this doesn't touch the ubuntu path, let me see what's going on there | 21:14 |
ianw | hmm, why has xenial stopped http://nb01.openstack.org/ubuntu-xenial-0000001395.log | 21:19 |
clarkb | fungi was the one debugging why tox-siblings works at all for us now and I think the determined state was we were using pip 9.x presumably due to older images | 21:20 |
ianw | clarkb: it's certainly true we don't seem to have built xenial in a few days, but it's not explicitly paused | 21:21 |
ianw | i think maybe it just can't get into build due to the other failing builds infront of it? | 21:21 |
clarkb | we did pause it on rax maybe that is global even if applied to a specific provider? | 21:21 |
ianw | oh these must be in emergency are they? | 21:22 |
ianw | oh no i just can't read; indeed pause: true next to xenial | 21:24 |
clarkb | ianw: ok commented on the dib change for pip10 support | 21:28 |
clarkb | ianw: I think the setuptools and rhel items should likely be addressed | 21:28 |
ianw | clarkb: yeah on tumbleweed, happy to do that but it wasn't there before, and this change i thought had enough stuffed into it | 21:30 |
clarkb | ya I didn't -1 for tumbleweed but for the others | 21:30 |
clarkb | we can sort out tumbleweed later | 21:30 |
clarkb | ianw: and I guess that leaves us with figuring out rax + xenial to open these floodgates | 21:31 |
ianw | ok ,will fix that up; thanks. the case statement was a late addition after it all started getting too complex, so yeah i dropped rhel accidentally, good catch | 21:31 |
ianw | yeah, ok i'll get back to that | 21:32 |
ianw | i spent almost a whole day trying to get something working on that friday with no luck | 21:33 |
clarkb | :/ | 21:33 |
clarkb | I'd offer to help but worried I've got so much of the pip specific stuff pages in at the moment to not be super useful | 21:33 |
clarkb | also bandersnatch upgrade so that we stop running out of disk on the mirror hopefully | 21:34 |
clarkb | ianw: re pinning on https://review.openstack.org/#/c/561691/ just use latest released version I guess? | 21:35 |
ianw | clarkb: looks like it's in 2.2.0 right? | 21:37 |
clarkb | yup | 21:38 |
ianw | up to you, i'd pin it because not 100% confident we monitor it closely enough to find problems that might pop us as they release | 21:39 |
clarkb | ++ I pinned | 21:39 |
*** vmlinuz has quit IRC | 21:45 | |
*** openstackgerrit has joined #openstack-dib | 21:52 | |
openstackgerrit | Nir Magnezi proposed openstack/diskimage-builder master: WIP: Do not use pip 10 or higher https://review.openstack.org/561740 | 21:52 |
*** andreas_s has joined #openstack-dib | 22:59 | |
*** andreas_s has quit IRC | 23:04 | |
openstackgerrit | Michael Johnson proposed openstack/diskimage-builder master: WIP: Do not use pip 10 or higher https://review.openstack.org/561740 | 23:09 |
clarkb | johnsom: instead of ^ can we work to get https://review.openstack.org/#/c/561479/ working? | 23:27 |
clarkb | johnsom: or at least do similar for ubuntu/debian as necessary? | 23:27 |
clarkb | I don't know that dib should be pinning pip10 for openstack specific reasons in its generic pip and virtualenv elements | 23:27 |
johnsom | clarkb Yeah, I have been poking at Ubuntu today. It's messy. ^^^ that, in my opinion is a short term fix while we get this other stuff figured out. | 23:30 |
clarkb | johnsom: also a link to failing dib build logs would likely be helpful | 23:31 |
clarkb | that chagne ahs a change id but from there i'm not sure what I'm supposed to look for | 23:31 |
johnsom | At least with ubuntu cloud-init requires python3-yaml so you are basically forcing things.... | 23:32 |
clarkb | johnsom: ya so one workaround is the one in the chagne I linked. Basically trick pip into thinking the old behavior is safe | 23:33 |
johnsom | clarkb Ah, sure. I was just fixing someone else's patch as I know they are signed off for the night. Give me a second and I will update the commit message | 23:33 |
clarkb | johnsom: another is to use virtualenvs exclusively (I've pushed a bunch of patches up to devstack, grenade, and ironic to explore that line of thought for those tools) | 23:33 |
clarkb | another would be to stop relying on system packages that have python deps | 23:33 |
clarkb | (use glean instead of cloud-init for example) | 23:34 |
*** rajinir has quit IRC | 23:34 | |
johnsom | clarkb Personally I like the venv path | 23:35 |
openstackgerrit | Michael Johnson proposed openstack/diskimage-builder master: WIP: Do not use pip 10 or higher https://review.openstack.org/561740 | 23:36 |
clarkb | in that particular case since its constrained to the amphora image wouldn't it be fairly straight forward to run that install in a virtualenv? | 23:37 |
clarkb | seems like just writing that change and moving forward would be the easiest thing | 23:38 |
johnsom | Yeah, fair point. Isn't going to solve the DIB issue though. It's the package-install/svc-map element that is broken, not our stuff. | 23:40 |
clarkb | dib isn't broken I don't think | 23:41 |
clarkb | at least not for that specific thing | 23:41 |
clarkb | johnsom: the thing that is failing is the amphora agent install beacuse it depends on pyyaml and pip10 won't let you remove the pyyaml from cloud-init right? | 23:42 |
clarkb | is something else broken too? | 23:42 |
johnsom | clarkb No, we don't ask for pyyaml | 23:42 |
johnsom | It's here: svc-map/package-installs.yaml:PyYAML: | 23:42 |
johnsom | The svc-map element in DIB | 23:42 |
johnsom | We use svc-map for redhat's alternate name.... | 23:43 |
clarkb | Collecting PyYAML===3.12 (from -c /opt/upper-constraints.txt (line 153)) | 23:43 |
clarkb | that happens in dib-run-parts Running /tmp/in_target.d/install.d/75-amphora-agent-install | 23:44 |
johnsom | clarkb https://github.com/openstack/diskimage-builder/blob/master/diskimage_builder/elements/svc-map/package-installs.yaml | 23:44 |
johnsom | Yeah, that element depends on svc-map | 23:44 |
clarkb | right but its not failing in preinstall.d | 23:44 |
clarkb | its failing in install.d | 23:45 |
johnsom | https://github.com/openstack/octavia/blob/master/elements/amphora-agent/element-deps | 23:45 |
johnsom | Feel free to grep octavia, we don't have it anywhere | 23:45 |
clarkb | the reason pyyaml may be in the image when amphora agent installs could be that element dep but I believe it would be there anyway for cloud init | 23:45 |
clarkb | the problem is amphora agent has a dep on pyyaml which can't be updated to the version in constraints | 23:45 |
johnsom | Sorry, no, we do not have a requirement on pyyaml | 23:46 |
clarkb | johnsom: its likely transitive | 23:47 |
johnsom | Yes, the image from ubuntu includes python3-yaml, if you include svc-map as a dep on your element, DIB fails out with pip 10 raising a conflict on PyYaml. | 23:47 |
johnsom | Let me comment that out and try a build | 23:47 |
clarkb | johnsom: I understand that what I'm trying to point out is its only failing because your software is also trying to install a nweer version of pyyaml | 23:47 |
clarkb | and pip10 doesn't like that for reasons | 23:47 |
clarkb | the fix here isn't to break dib | 23:47 |
clarkb | its to stop installing pip packages globally | 23:48 |
clarkb | unfortunately that decision was made for us by PyPa and pip and not something we really had any hand in making (in fact told them it would break us months ago, provided a workaround and were told to go away) | 23:49 |
clarkb | johnsom: oslo config is why you dep on pyyaml | 23:51 |
clarkb | (I just ran an install locally and if you leave of constraints it tells you what the dep comes from) | 23:51 |
clarkb | *leave off | 23:51 |
johnsom | Ah, another place. ok | 23:52 |
clarkb | and that is where it is failing in your job | 23:52 |
clarkb | pyyaml in svc map is system package | 23:52 |
clarkb | that works fine | 23:52 |
clarkb | what breaks is when you mix distutils installed system pacakges with globally installed pip packages | 23:52 |
clarkb | we can't control what system packges system packages dep on and we can't control that they use distutils and we can't control that pip10 fails when they conflict. But what we can do is stop installing pip packages globally | 23:53 |
johnsom | Yeah. Ok, somehow this got confused a bit. So, yeah, a venv for the amp would be easy enough and a good thing anyway. | 23:54 |
openstackgerrit | Michael Johnson proposed openstack/diskimage-builder master: WIP: Do not use pip 10 or higher https://review.openstack.org/561740 | 23:57 |
johnsom | Take the bad example case off that | 23:57 |
clarkb | I don't think its a bad example case | 23:58 |
clarkb | pip10 is causing that failure | 23:58 |
clarkb | I just disagree on the method being chosen to fix it :) | 23:58 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!