-
Notifications
You must be signed in to change notification settings - Fork 1.7k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Revert july9 prs #154
Revert july9 prs #154
Conversation
Hello. I am sorry for the trouble this causes you. The problem is expected although I did not know about OS template building directly from this repository. The problem is (as you can see in the error message) that newer Django (>=2.2) needs SQLite 3.8.3 but RHEL/Centos 7 have 3.7.17. The reason why I've updated this repository to Django 2.2 is that this version is LTS and all older versions 1.11 LTS, 2.0 and 2.1 are no longer supported. We were aware of this incompatibility for older RHEL/Centos so I've created a separated branch 1.11.x so users of RHEL/Centos 7 can still use the old 1.11.x version of Django. Would it be possible to:
|
@gabemontero is this only impacting the django template? (not the django-postgresql template which uses a postgresql image, not sqlite) samples operator doesn't appear to ship the pure django template, so we could be ok here if that's the case. (at least from a customer impact perspective). (we'd still have to decide what to do about the imageeco e2e test) |
@bparees the only template e2e's are failing on is The test is doing a We do not manually create that template and then instantiate it. It is a template that is installed. |
ok. I assumed since the error was on sqllite, that a deployment using psql would not have an issue, but i guess i'm wrong. So if we really want to do the "right" thing here for customers we need to:
|
OK, so we have this PR for that
So I believe this is the opposite of what @frenzymadness noted he did in #154 (comment) And this is necessary to support users on shipped versions of openshift ... namely 4.2 to 4.5 @bparees @frenzymadness does that make sense ... do you all understand and agree? If so, @frenzymadness , can you take on creating this new branch which includes the commits from the PRs from yesterday?
So this means in the new, "not master" branch in this repo, we want https://github.com/sclorg/django-ex/blob/master/openshift/templates/django-postgresql.json#L98 to point to the new branch but again not master branch .... we want the URL to changes such that "master" is the new branch name. That means that https://github.com/sclorg/django-ex/blob/master/openshift/templates/django-postgresql.json#L456-L460 would need to be updated (again, substitute the new branch name for master) to specify a default now ... namely our new branch Is that something you can take on @frenzymadness as well?
Presumably this is when 4.5 goes EOL, which would be either 2 or 3 years from now (don't remember off the top of my head. |
should be less than that. I believe 4.5 will EOL when 4.8 ships, so 9-12 months. Luckily this isn't in 4.6 since that's the extended support release (assuming we get this fixed up before 4.6 ships). see: https://access.redhat.com/support/policy/updates/openshift#dates |
yeah. we need master to work on the old version because master is what all the templates in the field are already pointing to. |
Although I try to understand where the problem is (I am sorry, I am not OpenShift user, just s2i-python-container maintainer) I don't like the idea that we'll show an obsoleted version with probably more obsoleted dependencies (with security vulnerabilities) to users coming to this repository looking for a good example. That's the idea why I've updated the master branch to supported versions and kept the older one harder to find. I can definitely switch the two branches so the new code will be in 2.2.x and the old one moved from 1.11.x to master. But only if there is no better solution. |
Side note: There were issues and PRs for updates years old so It'd be nice to maintain this example repository more proactively. There are also a lot of dead links to OS documentation. |
I agree i'm not thrilled with having to leave master in such a state either, but the sample is primarily consumed by openshift customers, so breaking it for all of them isn't a viable path forward. Again, switching the template to explicitly ref a non-master branch is a good solution to get us out of this eventually (when we feel a sufficient portion of our customers would no longer be using the old template). If there is another alternative, i'm certainly open to it.
Absolutely agreed. This sample is owned by the SCL team that owns the SCL images it is run on top of. Which it sounds like you're a part of as the s2i-python maintainer. I appreciate that you've gone through and addressed the long-standing idle PRs, that is what the SCL team was supposed to be doing all along. The reason @gabemontero and I are engaged is because we(openshift) run tests to confirm that those templates keep working, and in this case these changes broke those tests (Which means they broke the templates for our customers also). Ideally the SCL team (which also owns the template itself) is confirming they do not merge changes to the images, templates, or the examples the templates consume, that break them, though I understand that can be challenging to ensure, which is part of why we keep running the tests on the openshift side. There's no finger pointing here, just trying to find the healthiest way to move forward w/ the least impact to our users. |
My plan is:
Does that sound good to you? |
On Sat, Jul 11, 2020 at 2:17 AM frenzymadness ***@***.***> wrote:
My plan is:
1. Create a new branch 2.2.x from the current master branch
2. Reset current master branch to commit 6702b2b
<6702b2b>
(from Apr 14) which is the HEAD of 1.11.x branch now & force-push it.
3. Remove branch 1.11.x
4. Update notes in readmes in both branches to reflect actual state.
Does that sound good to you?
Yes - thanks.
… —
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#154 (comment)>, or
unsubscribe
<https://github.com/notifications/unsubscribe-auth/AA3NU5HCPBSD736XBN6RWKLR277XLANCNFSM4OV7MDJA>
.
|
merging a revert is generally preferred to rewriting git history, since rewriting history screws up anyone who's got a fork of this repo, or other branches. |
Is that enough to fix OpenShift troubles? Could you please take a look on #155 ? Is it safe to merge it? |
I started off a test job in openshift/cluster-samples-operator#300 @frenzymadness the e2e-aws-image-ecosystem will run the test we need to confirm things I personally do not know enough about django/python to know if #155 is safe based on just looking at it. But when I have a cluster up on Monday I'll see about hand crafting an OpenShift build that runs off of your PRs branch and let you know the results. |
thanks @frenzymadness @gabemontero ! |
@frenzymadness @lvarin
Since the combination of PRs
merged earlier today, we've started seeing docker builds from openshift templates defined in this repo start to fail.
I've posted the full build log in https://projects.engineering.redhat.com/browse/RHELPLAN-48796
It looks like there is a disconnect with the sqlite version that django expects. Perhaps there is a pending image update that is necessary?
Can you all report back with any analysis, including either determination of which commit(s) caused the disconnect, or what additional items are needed to get that build working again?
We can give it some time, but in the interim, any OpenShift customer attempting to use the template in question will encounter errors, so if fixing this gets too prolonged, we will to revert your recent changes until things can be sorted out.
thanks
/assign @bparees
@coreydaley FYI