Cheers on os.stg.fedoraproject.org!
I'd love to have waiverdb[1] be the first app on os.fedoraproject.org once it is ready. Do you guys have any estimates of when we could start messing with that in both stg and prod? (Maybe we could help you work out kinks if we can try deploying to os.stg.fp.o even before os.fp.o is ready?)
I put a due date in my project planning of "June 30th" for when waiverdb would be available in a production env. It's not "make or break" if we don't hit that date, but if os.fp.o won't be available for many months, maybe we should go ahead with VM-based ansible deployment in the meantime.
Yours- -Ralph
[1] https://fedoraproject.org/wiki/Infrastructure/Factory2/Focus/WaiverDB
On 05/15/2017 01:14 PM, Ralph Bean wrote:
Cheers on os.stg.fedoraproject.org!
I'd love to have waiverdb[1] be the first app on os.fedoraproject.org once it is ready. Do you guys have any estimates of when we could start messing with that in both stg and prod? (Maybe we could help you work out kinks if we can try deploying to os.stg.fp.o even before os.fp.o is ready?)
Well, it's pretty hard to estimate as we have so much going on and this is not the super top priority. That said, I think its not going to be too long. We have our staging instance up. We have successfully deployed an app from git and rebuilt it on changes. Next step we really need is to get a wildcard ssl cert so we can point people to the apps without going though a bunch of hoops (firefox won't show them to you because of HSTS, chrome/chromium you have to start with a insecure option). Hopefully that cert will happen soon and then the main thing left in my mind is to get apps setup in ansible. ie, ansible creates the project and such so we don't need to do any clicking around or cli stuff. We could probibly use help with that if anyone has used the openshift ansible modules. After that I could see trying to setup waverdb in stg.
I put a due date in my project planning of "June 30th" for when waiverdb would be available in a production env. It's not "make or break" if we don't hit that date, but if os.fp.o won't be available for many months, maybe we should go ahead with VM-based ansible deployment in the meantime.
I think that is doable.
kevin
Excerpts from Kevin Fenzi's message of 2017-05-17 14:34 -06:00:
Hopefully that cert will happen soon and then the main thing left in my mind is to get apps setup in ansible. ie, ansible creates the project and such so we don't need to do any clicking around or cli stuff. We could probibly use help with that if anyone has used the openshift ansible modules. After that I could see trying to setup waverdb in stg.
So I am very curious about this too, since we will want to use Ansible for managing our internal deployment of the apps as well.
Not sure which OpenShift Ansible module you are thinking of -- as far as I can see, nobody has actually written such a thing.
There is https://github.com/openshift/openshift-ansible but I think that is just roles and playbooks for deploying an OpenShift cluster itself (not for deploying apps inside the cluster).
I found this demo: http://servicesblog.redhat.com/2016/08/11/automating-the-provisioning-and-co... but they are just invoking the oc command using Ansible's command module. That's probably not a *bad* approach, if you just keep some OpenShift YAML checked into git alongside the playbooks, and just have the playbook do `oc apply -f`, although it doesn't give you any check mode or diff or anything nice.
Oh and I discovered there *is* a kubernetes module: https://docs.ansible.com/ansible/kubernetes_module.html ... but I'm not sure how compatible that would be with OpenShift?
On 05/22/2017 10:38 PM, Dan Callaghan wrote:
Excerpts from Kevin Fenzi's message of 2017-05-17 14:34 -06:00:
Hopefully that cert will happen soon and then the main thing left in my mind is to get apps setup in ansible. ie, ansible creates the project and such so we don't need to do any clicking around or cli stuff. We could probibly use help with that if anyone has used the openshift ansible modules. After that I could see trying to setup waverdb in stg.
So I am very curious about this too, since we will want to use Ansible for managing our internal deployment of the apps as well.
Not sure which OpenShift Ansible module you are thinking of -- as far as I can see, nobody has actually written such a thing.
Well, finally having had time to look more you are right.
I was thinking there were some modules in openshift-ansible for managing apps, but there doesn't seem to be.
However, there is: ansible-container that might be something we can leverage.
There is https://github.com/openshift/openshift-ansible but I think that is just roles and playbooks for deploying an OpenShift cluster itself (not for deploying apps inside the cluster).
Right. Even though it actually says 'manage' it does not (yet?) have any way to manage apps.
I found this demo: http://servicesblog.redhat.com/2016/08/11/automating-the-provisioning-and-co... but they are just invoking the oc command using Ansible's command module. That's probably not a *bad* approach, if you just keep some OpenShift YAML checked into git alongside the playbooks, and just have the playbook do `oc apply -f`, although it doesn't give you any check mode or diff or anything nice.
Oh and I discovered there *is* a kubernetes module: https://docs.ansible.com/ansible/kubernetes_module.html ... but I'm not sure how compatible that would be with OpenShift?
Yeah.
I think our best bet is to try and come up with some workflow around ansible-container. It currently also uses the oc commands in the end, but they are trying to move to the REST API from what I understand.
I am not sure at all what this workflow might be, going to take more looking into it before I could even propose something. It looks like ansible-container builds images locally, which I'd like to avoid. They should be built in the openshift pipeline.
Or if anyone sees any better ideas I am open to them.
I do want to avoid clicky or cli deployments that need a human to remember everything they clicked or typed. If we have to redeploy entirely, we should be able to do so from ansible.
Perhaps internal openshift folks have some thoughts here...
kevin
Adam pointed out:
https://github.com/openshift/openshift-tools
but I am still looking for the kind of management stuff I was hoping for there.
kevin
infrastructure@lists.fedoraproject.org