As mentioned in the last meeting, I've created a page to track what still needs to be migrated from puppet to ansible.
https://fedoraproject.org/wiki/Infrastructure/PuppetToAnsibleMigration
After a couple of false starts, I think I've gotten the right list of hosts to be migrated. I ended up using the 'if it's got an eth?_ip defined in host vars, then it's been migrated. That may not be a true assumption, but that's what I used to create the table.
I thought we should strike-through the hosts as they're done and maybe create a completed table later if it gets too unwieldy.
On Sat, Mar 01, 2014 at 11:06:30AM +1100, Graham Williamson wrote:
As mentioned in the last meeting, I've created a page to track what still needs to be migrated from puppet to ansible.
https://fedoraproject.org/wiki/Infrastructure/PuppetToAnsibleMigration
After a couple of false starts, I think I've gotten the right list of hosts to be migrated. I ended up using the 'if it's got an eth?_ip defined in host vars, then it's been migrated. That may not be a true assumption, but that's what I used to create the table.
Nice start!
There will likely be some overlap, but do you think we could generate a similar table for service rather than host. Like which applications remain to be port to ansible (bodhi is done, fedocal is done, fedmsg is done, glusterfs is done, but is mirrormanager?, fas?, koji?).
Just an idea :)
Thanks for the help! Pierre
On Sat, 1 Mar 2014 12:07:47 +0100 Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Sat, Mar 01, 2014 at 11:06:30AM +1100, Graham Williamson wrote:
As mentioned in the last meeting, I've created a page to track what still needs to be migrated from puppet to ansible.
https://fedoraproject.org/wiki/Infrastructure/PuppetToAnsibleMigration
After a couple of false starts, I think I've gotten the right list of hosts to be migrated. I ended up using the 'if it's got an eth?_ip defined in host vars, then it's been migrated. That may not be a true assumption, but that's what I used to create the table.
Nice start!
There will likely be some overlap, but do you think we could generate a similar table for service rather than host. Like which applications remain to be port to ansible (bodhi is done, fedocal is done, fedmsg is done, glusterfs is done, but is mirrormanager?, fas?, koji?).
Just an idea :)
Yeah, I agree that an application table would be good. ;)
Perhaps we can use https://apps.fedoraproject.org/ 's config to generate this list and then update from there?
Some hosts also we will NOT be migrating. In particular the app servers. Under our old model we had a pool of app servers that they ran every app. The new model we are moving to is that each app has it's own instances. This results in more instances at the end, but it means when one app is having issues it likely won't affect other apps and we can much more easily isolate what the problem is without the noise of all the other apps in the mix. So, we can stick a "N/A" on all the app servers.
A few other notes:
* openid* have been replaced by fedoauth* (already in ansible).
* bvirthost02/03 have nothing on them and are going to be repurposed, so they can be dropped from the list.
great stuff... thanks for working on it.
kevin
On Sat, 2014-03-01 at 11:17 -0700, Kevin Fenzi wrote:
On Sat, 1 Mar 2014 12:07:47 +0100 Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Sat, Mar 01, 2014 at 11:06:30AM +1100, Graham Williamson wrote:
There will likely be some overlap, but do you think we could generate a similar table for service rather than host. Like which applications remain to be port to ansible (bodhi is done, fedocal is done, fedmsg is done, glusterfs is done, but is mirrormanager?, fas?, koji?).
Some of these updated. I will be able to finish this off this afternoon after work.
Just an idea :)
Yeah, I agree that an application table would be good. ;)
Perhaps we can use https://apps.fedoraproject.org/ 's config to generate this list and then update from there?
Implemented.
Some hosts also we will NOT be migrating. In particular the app servers. Under our old model we had a pool of app servers that they ran every app. The new model we are moving to is that each app has it's own instances. This results in more instances at the end, but it means when one app is having issues it likely won't affect other apps and we can much more easily isolate what the problem is without the noise of all the other apps in the mix. So, we can stick a "N/A" on all the app servers.
A few other notes:
openid* have been replaced by fedoauth* (already in ansible).
bvirthost02/03 have nothing on them and are going to be repurposed, so they can be dropped from the list.
Changes updated.
great stuff... thanks for working on it.
kevin _______________________________________________ infrastructure mailing list infrastructure@lists.fedoraproject.org https://admin.fedoraproject.org/mailman/listinfo/infrastructure
On Mon, Mar 03, 2014 at 07:45:06AM +1100, Graham Williamson wrote:
On Sat, 2014-03-01 at 11:17 -0700, Kevin Fenzi wrote:
On Sat, 1 Mar 2014 12:07:47 +0100 Pierre-Yves Chibon pingou@pingoured.fr wrote:
On Sat, Mar 01, 2014 at 11:06:30AM +1100, Graham Williamson wrote: There will likely be some overlap, but do you think we could generate a similar table for service rather than host. Like which applications remain to be port to ansible (bodhi is done, fedocal is done, fedmsg is done, glusterfs is done, but is mirrormanager?, fas?, koji?).
Some of these updated. I will be able to finish this off this afternoon after work.
Cool, thanks for making the list.
I removed bugzilla since this is not something we have the hand and is thus neither in ansible nor puppet.
I also striked fedocal, nuancier and FMN (notifications) which I know are already in ansible.
Pierre
infrastructure@lists.fedoraproject.org