| Moha_ | I have installed kolla-ansible version 18.6.0 (to match to the production which has ben deployed 1 year ago). I have save images from production env and loaded to docker per nodes. For example I have quay.io/openstack.kolla/cron:18.6.0 on all nodes. The pull step FAILED, saying cant find this version. I checked from the browser: https://quay.io/repository/openstack.kolla/cron?tab=tags and, yes, there's no tag 18.6.0 there? | 00:33 |
|---|---|---|
| Moha_ | Why?! | 00:33 |
| Moha_ | (in my gloabals.yml, openstack version has set to 18.6.0 to be compatible with our production env) | 00:34 |
| Moha_ | Part of the Error: | 00:37 |
| Moha_ | https://www.irccloud.com/pastebin/pdg74RwL | 00:37 |
| Moha_ | Loaded images to gnocchi server (as an example): | 00:38 |
| Moha_ | https://www.irccloud.com/pastebin/XQ2SejO6 | 00:39 |
| Moha_ | Are 18.6.0 removed from quay? | 00:40 |
| Moha_ | How can I say the kolla not to pull from the registry stuff while they exist on the nodes, any settings for globals.yml? | 00:42 |
| mnasiadka | Moha_: no, they were never there - we have different tag naming in quay.io and you should not use these container images in production - you should build your own | 07:02 |
| Moha_ | But on our production env, I see images are tagged with 18.6.0?! And also `openstack_release: "18.6.0"` in our production globals.yml | 07:10 |
| Moha_ | Then how it's been deployed with 18.6.0 (which is a minor update inside 2024.1)?! 🤔 | 07:12 |
| Moha_ | mnasiadka: By 'build your own', you mean setting `kolla_install_type: "source"` in globals.yml file? | 07:15 |
| mnasiadka | Moha_: no, by build your own I mean using kolla-build to build them and push into a registry - https://docs.openstack.org/kolla/latest/admin/image-building.html | 07:18 |
| mnasiadka | In your production env either the quay.io images have been retagged to 18.6.0 (I doubt it) or built with kolla-build and pushed into some registry | 07:19 |
| Moha_ | Ah I see. Thanks for the point. | 07:50 |
| opendevreview | Taavi Ansper proposed openstack/kolla-ansible master: Fix keystone with IDP configured. https://review.opendev.org/c/openstack/kolla-ansible/+/975901 | 08:14 |
| opendevreview | Taavi Ansper proposed openstack/kolla-ansible master: Fix keystone with IDP configured. https://review.opendev.org/c/openstack/kolla-ansible/+/975901 | 09:00 |
| *** jhorstmann is now known as Guest3432 | 10:09 | |
| tafkamax | ugh I hate from_json | 10:37 |
| tafkamax | It requires the data it loads in to be "proper" json | 10:37 |
| tafkamax | the attribute_mapping example for federation mappings is a list of dicts | 10:37 |
| tafkamax | But for json that is a no-no | 10:37 |
| tafkamax | it needs to start as an object not as an list | 10:37 |
| tafkamax | hmm or maybe not... maybe there is another issue at hand... | 10:40 |
| tafkamax | running a normal "debug" operation loads in | 10:44 |
| tafkamax | but kolla-toolbox fails | 10:44 |
| tafkamax | maybe this is an issue "whre is the file located" ? | 10:45 |
| tafkamax | will try to wrap kolla-toolbox over the debug command | 10:45 |
| tafkamax | Really :S issue | 11:27 |
| tafkamax | "msg": "dictionary requested, could not parse JSON or key=value"} | 11:27 |
| tafkamax | When trying to load in a file -> to ansible vars and then insert it here as rules: https://docs.ansible.com/projects/ansible/latest/collections/openstack/cloud/federation_mapping_module.html | 11:28 |
| Moha_ | Instead of `kolla-ansible deploy`, how to split deployment in a right ordeer? I mean: | 11:30 |
| Moha_ | first: `--limit controller-1,controller-2,controller-3`, second: `--limit compute-1,compute-2,compute-3`, third: etc | 11:31 |
| Moha_ | 1️⃣ Deploy Controllers Only > 2️⃣ Deploy Compute Nodes Only > 3️⃣ Deploy Observability / Prometheus / Ceilometer Only > etc? | 11:32 |
| Moha_ | or deploying service by service: mariadb,keystone > glance > nova > prometheus > ceilometer, using --tags | 11:34 |
| Moha_ | Splitting 'deploy' helps to handle errors better. Most of the time, after resolving a fatal issue, I'm not sure how to re-run the script, from the scratch by `kolla-ansible depeloy` or limiting it to some nodes. How do you encounter with this situation? | 11:36 |
| tafkamax | I guess it depends on the service | 11:46 |
| tafkamax | First you want your core services to be on only | 11:47 |
| tafkamax | e.g. mariadb, lb, rmq | 11:47 |
| tafkamax | s/only/online*/ | 11:47 |
| tafkamax | s/on only/online\*/ | 11:47 |
| tafkamax | Supeerr weird issue, can not get the openstack.cloud to work properly with the rules. :( | 12:37 |
| tafkamax | feels bad, battling stupid ansible issues | 12:37 |
| tafkamax | Will try to use the old command version for creating attribute maps, lets see if it helps for now. | 12:39 |
| tafkamax | If any1 has and ideas, I will be happy to hear your thoughts. | 12:40 |
| tafkamax | I have a feeling this is a problem openstack.cloud.federation_mapping module | 13:08 |
| tafkamax | as I am using the example config from the module and it simply does not work | 13:08 |
| tafkamax | Hmm maybe sdk_log_level and sdk_log_path will help | 13:11 |
| tafkamax | hmm those didnt help... | 13:14 |
| tafkamax | AI overlord is saying it might be kolla-toolbox issue hmm | 13:25 |
| tafkamax | mnasiadka: I found out the openstack.cloud.federation_mapping issue is related to kolla_toolbox. Using the module on a localhost without the wrapper works. | 13:36 |
| tafkamax | Do you know this is a known issue? | 13:36 |
| tafkamax | https://github.com/openstack/kolla-ansible/blob/master/ansible/library/kolla_toolbox.py#L125 | 13:42 |
| tafkamax | It is due to reading in the variables to json and then converting them back to variables for ansible, which seems to break list objects. | 14:23 |
| tafkamax | pretty brutal oversight tbh | 14:23 |
| tafkamax | And amazed that no1 else as stumbled onto this | 14:23 |
| tafkamax | or there is a fix that i just dont know :) | 14:23 |
| opendevreview | Taavi Ansper proposed openstack/kolla-ansible master: Fix keystone with IDP configured. https://review.opendev.org/c/openstack/kolla-ansible/+/975901 | 14:53 |
| tafkamax | If this gets fixed I am hoping that is the final thing in my road to fix a simple but that botches many deployments. | 14:55 |
| tafkamax | I don't unfortunately have a good idea how to fix the kolla_toolbox issue yet. AI doesn't seem to fully help aswell. | 14:55 |
| tafkamax | the other openstack.cloud modules seem to work nicely in the federation configuration process, which is good | 15:08 |
| opendevreview | Taavi Ansper proposed openstack/kolla-ansible master: Fix keystone with IDP configured. https://review.opendev.org/c/openstack/kolla-ansible/+/975901 | 15:13 |
| opendevreview | Taavi Ansper proposed openstack/kolla-ansible master: Fix keystone with IDP configured. https://review.opendev.org/c/openstack/kolla-ansible/+/975901 | 15:57 |
| tafkamax | Won't deal with this any more today. Maybe somebody has some thoughts on how to tackle the lists problem. | 16:38 |
| tafkamax | * lists problem with kolla-toolbox. | 16:38 |
| tafkamax | Changing the kolla-toolbox code wasn't very helpful in the ci | 18:11 |
| tafkamax | Maybe one way is to just revert to the old command: module for the attribute mapping and not frustrate myself with kolla-toolbox, but that would be a half assed solution :/ | 18:22 |
| Moha_ | Do you recommend migrating from docker to podman? (What are the advantages?) or it doesn't matter? | 21:35 |
Generated by irclog2html.py 4.0.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!