Closed Rezney closed 6 months ago
@swapdisk @heatmiser Please let me know if the approach is ok.
Couple of questions:
analysis_repos_el7
from "playbooks/analysis.yml" is set to copr:copr.fedorainfracloud.org:group_oamg:leapp
to enable the repo but there is nothing enabling the Fedora Copr. Do we want to introduce it or there are other plans?leapp_preupg_opts
from "playbooks/analysis.yml" seems to be overriding leapp_upgrade_opts
from the defaults. What about introducing a list and extending it? Please let me know if the approach is ok...
Yes, I like this approach.
analysis_repos_el7
from "playbooks/analysis.yml" is set tocopr:copr.fedorainfracloud.org:group_oamg:leapp
to enable the repo but there is nothing enabling the Fedora Copr. Do we want to introduce it or there are other plans?
Right, I don't know what's going on with that. This is just an example playbook, but it still should make sense. Anybody?
leapp_preupg_opts
from "playbooks/analysis.yml" seems to be overridingleapp_upgrade_opts
from the defaults. What about introducing a list and extending it?
Sounds good if we can do it in a forward compatible way, i.e., not breaking existing playbooks folks may have that are already overriding leapp_upgrade_opts
?
I believe the copr fedora stuff was an example but I am not positive @jeffmcutter do you remember?
That was added after my time:
https://github.com/redhat-cop/infra.leapp/commit/7a647c8e2b13a9db8288a50e5d399e4b97bb4c4b
@ingmarfjolla Can you help shed some light on the question?
If I remember correctly at the time I was adding those examples, we were pointing to copr:copr.fedorainfracloud.org:group_oamg:leapp
for 2 reasons. 1 was to just show the usage of the analysis_repos_el7
variable in general, and 2 we were testing an upstream fix for an issue we had with leapp at the time (I don't work at RH anymore but in the leapp gchat there should still be a a thread about it). Heres the original issue where I made the suggestion (and makes mention of the issue we ran into and tested)
Thanks all for chiming in. I wanted to make sure I am not stepping on anyone's toes with enabling it. I believe it is really valuable for testing the upstream bits as @ingmarfjolla mentioned.
Oh I remember it was because the bug fix for xfs wasn't downstream yet
@Rezney it is my opinion that we move away from requiring the end user to manually set leapp_upgrade_type
and then rely on logic that stems from that setting. Instead, I feel that we should be programmatically introspecting each system and then setting the requisite repos that need to be enabled. I began this type of effort in the leapp-project we use for workshops, here. We should be able to determine where/how a particular node/instance is running, review either current subscription mode or currently enabled repos to then set the appropriate repos needed for leapp execution. In this manner, we could have 5 different RHEL nodes in a particular inventory for an ansible leapp playbook, with some running on bare-metal, some on VMware, some on various clouds, subscribed viu Satellite or RH CDN or via RHUI...each node/instance could be handled correctly.
@heatmiser Nice! I was not aware we have "Amazon EC2" info for free and I counted with the natural reserve against bringing in more dependencies to get e.g. "ec2_metadata_facts". Do you happen to know we have similar info for Azure and GCP (lazy ask - I can check) ? I definitely agree with the diverse inventory reasoning. Ok, let's go this way.
how is this coming along or are you still waiting for some answers?
I plan to resume working on this next week. We have been snowed under because of a new release. Sorry for the delay.
how is this coming along or are you still waiting for some answers?
I have been working on this intermittently over the past month. I am planning to submit a PR over the weekend.
how is this coming along or are you still waiting for some answers?
I have been working on this intermittently over the past month. I am planning to submit a PR over the weekend.
Sounds good to me. Closing this PR then.
Support base os PAYG instances on AWS, Azure, and GCP by adding:
Technically, we could use e.g. "ec2_metadata_facts" but we would have to add new dependencies.