*** rfolco has joined #softwarefactory | 01:10 | |
*** rfolco has quit IRC | 01:28 | |
*** zenkuro has joined #softwarefactory | 05:24 | |
*** zenkuro has quit IRC | 06:32 | |
*** zenkuro has joined #softwarefactory | 06:33 | |
lyr | thanks for the example playbook tristanC | 07:29 |
---|---|---|
lyr | About certificates : lecm seems to have been removed, both the sf-lecm ansible role, and I found no reference in the doc about lecm or letsencrypt. But the network.use_letsencrypt is still here. Should we deal with this ourselves ? | 07:33 |
lyr | Or SF have a replacement for lecm ? | 07:33 |
*** apevec has joined #softwarefactory | 08:16 | |
zenkuro | hi! Is there a place where I can read more about concept of tenants? Ive checked out https://softwarefactory-project.io/docs/guides/tenant_deployment.html?highlight=resources#workflow-details and https://softwarefactory-project.io/docs/guides/unmanaged_tenant.html?highlight=resources but it is not enough for me | 08:36 |
*** rfolco has joined #softwarefactory | 09:03 | |
*** rfolco has quit IRC | 09:07 | |
*** rfolco has joined #softwarefactory | 09:51 | |
lyr | I think "A tenant SF is an instance that does not run Zuul services" says it all ? | 09:54 |
zenkuro | lyr: may be Im missing something but what is the point of SF without zuul? + this tenant can run jobs and upload logs | 10:05 |
lyr | zenkuro: maybe managesf + gerrit on an instance, zuul & nodepool on another instance, ... ? | 10:11 |
zenkuro | lyr: hm, does it help with security or resource allocation? For now I can see that different tenant can be used in combinations with different acl and different base job | 10:13 |
zenkuro | from perspective of code organization. | 10:13 |
zenkuro | or for instance to minimize damage to main config repo if you want to experiment with things | 10:15 |
lyr | I don't know | 10:18 |
lyr | I'm a beginner in the topic | 10:18 |
lyr | Just trying to get nodepool to build an openstack based image atm | 10:19 |
zenkuro | lyr: lol, Ive havent get to nodepool yet, Im trying to figure out how to manage logs with static machines =) | 10:21 |
lyr | Well,trying to build debian image, requies debootstrap, which is available in EPEL only, and software factory is incompatible with EPEL | 10:28 |
zenkuro | lyr: m... hm... that is strange to hear | 10:49 |
zenkuro | lyr: Im curious on how you gona solve this? like create a dedicated machine and ssh to it to build there? is it possible with SF? | 11:15 |
zenkuro | or make it as a job? | 11:16 |
lyr | for now the ugly way | 11:31 |
lyr | removed sfconfig rpm, installed epel, installed debootstrap, removed epel, installed back sfconfig | 11:32 |
lyr | My own "install SF" config will do that as an initial step, before installing software factory | 11:32 |
lyr | But I'm waiting for tristanC or other knowledgeable guy to solve this and other issues | 11:33 |
zenkuro | wow... that is dangerous, it is better to do snapshots if you can | 11:56 |
tristanC | lyr: oops, i thought we did a release note about droping lecm. You do need to manage tls yourself | 12:31 |
tristanC | zenkuro: the point of sf tenant is to isolate the service that do not have tenancy, for example gerrit, logserver or kibana | 12:31 |
tristanC | zenkuro: to manage different zuul config/resources, you can just create zuul tenant, no need to deploy another instance | 12:32 |
tristanC | lyr: you don't have to use disk-image-builder, you can use the virt-customize role to customize an existing debian cloud image | 12:33 |
zenkuro | tristanC: I was kinda experimenting with it now and I stopped at at getting check job show up https://softwarefactory-project.io/docs/guides/unmanaged_tenant.html | 12:34 |
tristanC | zenkuro: yeah, that seems like the documentation you are looking for | 12:35 |
tristanC | zenkuro: and here is the zuul documentation to control the acl per zuul tenant: https://zuul-ci.org/docs/zuul/reference/tenants.html#attr-tenant.allowed-triggers | 12:36 |
zenkuro | tristanC: many thanks! | 12:37 |
tristanC | zenkuro: you're welcome :) | 12:39 |
zenkuro | but may be you can enlight me regarding the reason that made me investigate this thing: report-build-page | 12:39 |
zenkuro | sorry for throwing one question after another | 12:39 |
tristanC | zenkuro: no worries, what is the question? | 12:39 |
zenkuro | it looks like this property is auto generated: resources/_internal.yaml: zuul/report-build-page: True | 12:40 |
zenkuro | so when I change it at tenant: local definition around my project it has no effect | 12:41 |
zenkuro | so I was thinking if I should create a separate tenant in order not to mess up with auto-generated configs | 12:42 |
tristanC | right, this is a tenant level configuration, and it's hardcoded to true at the moment, so what you are doing is correct: create a separate tenant with your custom configuration | 12:43 |
tristanC | actually we would like to rename the default `local` tenant to `internal`, and make it manage the config update (e.g. nodepool and zuul top-level tenant configuration) | 12:44 |
tristanC | because it's quite tricky to update the `local` tenant when needed (e.g. when the config-update job change) while keeping user configuration | 12:45 |
zenkuro | tristanC: sounds cool, it would be nice if you also deliver pre build `external` tenant, since I have issues with making `check` job from the guide to appear | 12:46 |
zenkuro | tristanC: false alarm! sfconfig + reboot saves the day =) | 12:46 |
tristanC | yeah that would be an useful tool to have, i agree it's a pain to setup a new tenant from scratch | 12:47 |
*** rfolco has quit IRC | 13:22 | |
*** rfolco has joined #softwarefactory | 13:22 | |
zenkuro | by the way is there a support of zuul jobs? Im mentioning this because turning them on by upstream_zuul_jobs: true leads to: | 13:58 |
zenkuro | [ara-report : Create the ARA database report directory] {"msg": "There was an issue creating ara-report as requested: [Errno 30] Read-only file system: b'ara-report'", "path": "ara-report"} | 13:59 |
zenkuro | just if somebody is interested in this thing. This happens on config check. So it should not be a consequences of my "experimentation" | 14:00 |
lyr | tristanC: well, the big idea behing using dib was to use the elements provided to get a basic debian buster image. We don't need any tweaking, so I expected nodepool's vanilla setup to speed up things | 14:06 |
tristanC | lyr: dib can be used to produce special image, it is used in opendev to setup a custom cloud-init and install zuul ssh key. It is quite a complicated process since it creates the image from scratch. | 14:34 |
tristanC | lyr: if you don't need dib elements, then it may be easier to just add the zuul ssh key on an image provided by the distro, it will be closer to what your user use | 14:35 |
lyr | I need the easiest way possible | 15:24 |
lyr | Not in a complicated process way of | 15:25 |
lyr | I mean | 15:25 |
lyr | I don't care if you're doing complicated stuff or not | 15:25 |
lyr | I care about spending the least amount of time on the topic | 15:25 |
lyr | Otherwise I'ld go for a packer something, since that's my prefered way of dealing with those question | 15:25 |
zenkuro | lyr: common it is open-source and project is under dev | 15:48 |
lyr | zenkuro: I know. As a new user I'm giving my input about adopting the project | 15:53 |
tristanC | lyr: have you seen https://softwarefactory-project.io/cgit/software-factory/sf-config/tree/ansible/roles/sf-repos/files/config/nodepool/virt_images/README.md ? | 15:59 |
lyr | yes | 16:00 |
tristanC | lyr: if that's too complicated, then you can also use you own tool to create the disk image, and uses the `cloud-image` nodepool configuration | 16:01 |
lyr | I'm trying to get https://paste.garrigue.re/?b04e8cda9d62a8dd#6HE4NpYSHBXRLV8KncjtYGrSmfqkQ2Dmgtfa2cWXWYwA to work | 16:02 |
lyr | I got the image built, my current issue being about uploading, but that's most likely about our openstack connection | 16:03 |
lyr | (then ofc try to get a job running with the image) | 16:03 |
tristanC | lyr: i think you need other elements to add the ssh key and setup ssh, have you seen opendev dib config https://opendev.org/openstack/project-config/src/branch/master/nodepool/nb03.opendev.org.yaml ? | 16:04 |
lyr | I sort of expected Zuul / Nodepool to deal with that | 16:08 |
lyr | Like adding one of /var/lib/software-factory/bootstrap-data/ssh_keys/*.pub | 16:08 |
lyr | In one of the debian elements' requirements in this list | 16:09 |
lyr | Expanded element dependencies to: dpkg sysprep modprobe runtime-ssh-host-keys base install-static debootstrap debian install-bin dib-python package-installs pkg-map dib-init-system debian-minimal manifests install-types openssh-server | 16:09 |
tristanC | lyr: Zuul / Nodepool doesn't know how to deal with that, you need to use an element such as nodepool-base | 16:09 |
lyr | I guess I need to add zuul-worker-user & nodepool-minimal from /usr/share/sf-elements ? | 16:10 |
tristanC | lyr: yes | 16:11 |
*** csomh has quit IRC | 16:44 | |
lyr | Was there any deprecation between 3.3 & 3.5 nodepool version regarding the glance api version ? | 16:52 |
lyr | We've an ageing openstack running glance api v1 | 16:53 |
lyr | I guess it's the reason why nodepool can't upload images, while our legacy SF 3.3 was ok | 16:54 |
lyr | I see nothing in clouds.yaml to enforce a legacy version usage | 16:55 |
tristanC | lyr: sf-3.5 integrate a more recent nodepool that requires a new openstacksdk. you should be able to pin the api version in the clouds.yaml | 17:02 |
lyr | tristanC: something like clouds.openstack.image_api_version: 2 ? | 17:04 |
tristanC | lyr: i guess yeah | 17:05 |
lyr | hmpf... hardstuck on this one, got no lead | 17:23 |
lyr | nodepool-builder.log is full of https://paste.garrigue.re/?50ebc5e96cb62342#5F5DLNmwjGjKkHZMogPQC3JrAWhHYGREMdmGr5h82RXs | 17:24 |
tristanC | lyr: perhaps add your couds.yaml to the paste and ping mordred about it? | 17:31 |
*** apevec has quit IRC | 17:32 | |
*** apevec has joined #softwarefactory | 17:33 | |
lyr | done | 17:40 |
*** zenkuro has quit IRC | 21:39 | |
*** apevec has quit IRC | 21:46 | |
*** rfolco has quit IRC | 23:06 | |
*** rfolco has joined #softwarefactory | 23:21 | |
*** rfolco has quit IRC | 23:53 | |
*** rfolco has joined #softwarefactory | 23:53 | |
*** rfolco has quit IRC | 23:58 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!