| *** ijolliffe has quit IRC | 01:32 | |
| *** TxGirlGeek has quit IRC | 01:49 | |
| *** HuyNguyen has joined #starlingx | 02:37 | |
| *** hyunsikyang has joined #starlingx | 05:16 | |
| *** hyunsikyang has quit IRC | 09:04 | |
| *** hyunsikyang has joined #starlingx | 10:36 | |
| *** ijolliffe has joined #starlingx | 13:47 | |
| *** HuyNguyen has quit IRC | 13:51 | |
| *** HuyNguyen has joined #starlingx | 13:51 | |
| *** HuyNguyen has quit IRC | 13:54 | |
| *** HuyNguyen has joined #starlingx | 13:55 | |
| *** HuyNguyen has quit IRC | 14:21 | |
| *** HuyNguyen has joined #starlingx | 14:21 | |
| *** TxGirlGeek has joined #starlingx | 17:40 | |
| sgw | dpenney: question about the tagging, we did not talk about it in the build meeting. | 18:09 |
|---|---|---|
| sgw | Could this cause a helm issue during the running of the ansible playbook? | 18:09 |
| sgw | dpenney: further Q for my understanding, if we have PBR, would this help solve this issue since an stx-3.0 fix could get a v3.y.z tag while a stx-4.0 fix would be v4.x.x? | 18:20 |
| dpenney | PBR wouldn't affect this. This is to give us a mechanism to lock down static image references in places like the ansible playbook that are currently referencing master-centos-stable-latest at runtime, which would basically be a moving target | 18:41 |
| dpenney | in the case of the bug I referenced, it's running the 2.0 load, but referencing master-centos-stable-latest for one of the images, which picks up an incompatible image | 18:41 |
| sgw | dpenney: thanks, ignore the helm issue, that was a proxy issue on my end | 18:54 |
| dpenney | images like the openstack ones that get pulled in by a built/managed application are fine, as they get built referencing the versions from the loadbuild. But once we added some starlingx images into installed charts/whatever with static references, we needed some way to have a locked down version ref | 18:57 |
| sgw | dpenney: do you need to lock down both openstack images and starlingx images? Why can't PBR be used with the starlingx genrated images? It seems that we are then creating another versioning/tagging system | 19:00 |
| dpenney | the openstack images get locked down by referencing the image records. CENGN builds loads weekly with a generated version tag based on timestamp, and the application build uses the record of those builds to update the charts with references to those versions. | 19:02 |
| dpenney | ie. CENGN runs build-stx-images and specifies a version label like 20200128T000001Z.0, which gets added into the image tag of master-centos-stable-20200128T000001Z.0 | 19:03 |
| dpenney | and all images that are built in that run get that same timestamped tag | 19:04 |
| dpenney | and then retagged with -latest | 19:04 |
| dpenney | there is currently no independent versioning for individual images, and nothing to say "get the version for this image from repo X's PBR" | 19:05 |
| dpenney | multiple ways for the images to get built, multiple ways to get the content used for building the image | 19:05 |
| sgw | dpenney: so could we change this to create a method to get the version from the repo X's PBR? | 19:16 |
| sgw | or more specificaly package Y's PBR | 19:17 |
| dpenney | seems like it would be complicated based on the wide variety of ways the images can be built, and would also require some build avoidance and load management. ie. building image X which is primarily using package Y, so use Y's PBR as the version for the image... so if Y hasn't changed, don't rebuild it... but maybe dependencies or other components have changed, so maybe rebuild it, but versions would collide | 19:21 |
| dpenney | since these images are more than just the primary package | 19:22 |
| sgw | There must be a way to simplfy this problem, it seems overly complex | 19:48 |
| *** TxGirlGeek has quit IRC | 20:05 | |
| sgw | slittle1: I am trying the distro layer and it seems to not find the compile layer's rpms, I have some logs with missing files | 21:00 |
| slittle1 | trying ... | 21:05 |
| *** TxGirlGeek has joined #starlingx | 21:07 | |
| slittle1 | which step is complaining ? | 21:09 |
| slittle1 | sgw: which command is failing ? | 21:12 |
| sgw | download_mirrors.sh | 21:12 |
| slittle1 | I think it's a caching problem | 21:24 |
| slittle1 | did you see the 'run yum --enablerepo=CENGN_Starlingx-cengn_compiler_installer_layer clean metadata' in the logs | 21:24 |
| sgw | I do now! | 21:24 |
| sgw | I am running as non-root, that command fails | 21:25 |
| sgw | I am also struggling with the cento8 mirror, a different issue, but today is a bad mirror day. | 21:28 |
| slittle1 | sgw: 'yum clean metadata' didn't clear it up for me ... checking id repodata is up to date | 21:30 |
| sgw | Ah so your seeing a similar problem? | 21:30 |
| slittle1 | ok I understand it | 21:43 |
| slittle1 | cengn's miror is now trying to mirror cengn's own build | 21:44 |
| slittle1 | download mirror is mapping mirror/starlingx/master/centos/compiler/latest_build/outputs/RPMS/std to mirror/centos/mirror.starlingx.cengn.ca/mirror/starlingx/master/centos/compiler/latest_build/outputs/RPMS/std | 21:45 |
| slittle1 | it doesn't mirror rpm.lst for one | 21:48 |
| slittle1 | and the mirror's repodata is out of date | 21:50 |
| *** sgw has quit IRC | 21:54 | |
| *** sgw has joined #starlingx | 22:11 | |
| *** ijolliffe has quit IRC | 22:18 | |
| sgw | slittle1: any update? | 22:28 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!