=== tmclaugh[work]_ is now known as tmclaugh[work]
=== tmclaugh[work]_ is now known as tmclaugh[work]
=== natorious is now known as zz_natorious
=== rangerpbzzzz is now known as rangerpb
[14:02] <larsks> harlowja: btw, submitted that pr yesterday.  And now it has tests, too.
=== rangerpb is now known as rangerpbzzzz
=== zz_natorious is now known as natorious
=== natorious is now known as zz_natorious
[15:39] <Odd_Bloke> smoser: I'm looking at fixing https://bugs.launchpad.net/ubuntu/+source/cloud-init/+bug/1411582 in cloud-init.
[15:40] <Odd_Bloke> smoser: My suggested fix is to shift the udev rule that creates /dev/disk/azure/... in to cloud-init, and then teach cloud-init to dereference /dev/disk/... symlinks.
[15:41] <Odd_Bloke> smoser: We could also look at making the directory created more generic (e.g. /dev/disk/cloud-init/...), so that we have a single place to look for stuff on different clouds.
[15:42] <Odd_Bloke> (This would, for example, allow us to answer 'do I have an ephemeral disk?' by just looking at /dev/disk/cloud-init/ephemeral or whatever)
[15:42] <Odd_Bloke> smoser: Thoughts?
[17:02] <smoser> Odd_Bloke, well, looking forward, there is a plan for a block device usage languange. that explicitly declares how disks should be used. this will support raid, lvm, bcache ... 
[17:02] <smoser> that is happening for curtin usage.
[17:02] <smoser> i'd like for cloud-init to be able to take that language and render it into being also.
[17:07] <smoser> the added piece that cloud-init needs is kind of what you were mentioning. curtin will be fed disk config like http://paste.ubuntu.com/10939715/ .
[17:08] <smoser> but ideally cloud-init can then allow substitution of things like 'ephemeral0' where it finds out what that is.
[17:23] <harlowja> larsks awesome
[17:23] <Odd_Bloke> smoser: Right, is that something that can happen now?  We've got pressure from Azure to get this fixed.
[17:24] <Odd_Bloke> smoser: (Also, will we realistically ever take the curtin language in to 0.7.x?)
[17:26] <smoser> Odd_Bloke, probably not
[18:00] <smoser> Odd_Bloke, so i think i'm ok with the overall idea.
[18:00] <smoser> then cloud-init would allow users to specify 'ephemeral0' as they can elsewhere.
[18:00] <smoser> and on azure that would map to /dev/cloud-init/ephemeral0
[18:00] <smoser> i think thats what you're saying
[18:16] <Odd_Bloke> smoser: Yeah, it would be /dev/disk/cloud-init/ci-ephemeral or somesuch.
[18:16] <Odd_Bloke> smoser: Ben proposed this udev config: http://paste.ubuntu.com/11566987/
[18:17] <smoser> it doesnt seem unreasonble.
[18:18] <smoser> but i dont know how this determines between "running on azure" and "running on hyper-v"
[18:19] <alexpilotti> smoser: this is for cirros?
[18:19] <smoser> alexpilotti, that above was azure ubuntu
[18:20] <alexpilotti> oh, k tx
[18:20] <Odd_Bloke> smoser: I'm not sure either; udev is not my strong suit. Let me check with Ben.
[18:23] <Odd_Bloke> smoser: Ben confirmed with Microsoft that that would only match in Azure environments.
[18:27] <smoser> why would they only match in azure
[18:27] <smoser> that is odd
[18:27] <smoser> dont you think ?
[18:27] <smoser> ID_VENDOR of Msft
[18:27] <smoser> odd
[18:35] <Odd_Bloke> smoser: I _think_ it's the UUIDs that are Azure-specific.
[18:36] <smoser> ah. that does seem more reaosnable.
[21:16] <alexpilotti> smoser: I’m hearing about ext4 for config drives
[21:48] <smoser> alexpilotti, that is odd / roughly stupid
[21:49] <smoser> not really here. but a link to info woudl be good. but really cloud-init doesnt care. 
[23:35] <harlowja> windows dll fun scares me lol