=== matsubara is now known as matsubara-afk [13:28] anyone here have a link or know what environment variables I can feed juju? [13:28] also is it possible to choose which server to add a charm to? === realnorth_ is now known as realnorth [13:31] realnorth, i have been told that jitsu has a -deploy-to option [13:31] but i never try it yet [13:33] i'm guessing just about everyone is sleeping, lifeless and bigjools at least. allenap maybe around? [13:33] i hadnt' followed all the dns thread [13:33] smoser: Hi there, what's up? [13:33] but i thought that lifeless suggested doing ec2 like names, which i think is generally useful. [13:34] but then others didn't like that because of "no human friendly hostname" [13:34] wouldn't cnames fix that? [13:34] and having dhcp hand out the hostname as the human-friendly-cname? [13:34] smoser: Yes. One of the questions I had was: what do names with the IP address in add beyond just using IP addresses? [13:35] rvba: ^ maybe you can help too? [13:35] not much i guess. [13:35] but if you add in the cname, then you have both, dont you? [13:35] Yeah. [13:36] For smaller clouds, having a nice name is more useful. For massive clouds it's a burden. [13:37] random bit of info . i think that there was some dns service that just converted "v.x.y.z.somedomain.com' to v.x.y.z Ip address for you [13:37] i tihnk ii think you'd just allow it as configurable. [13:37] an dif someone sets the "hostname " for something then they get it. [13:38] who cares. [13:38] smoser: so a node would have an ec2 like name and a "human" CNAME then? [13:38] yeah. [13:38] well, that was my thought. [13:38] reverse looking up its IP would get the ec2-like name [13:38] but who cares. [13:38] right [13:39] The main problem with writing DNS config files is with reverse config files (because of the classful/classless pb) so yeah, that sounds like something we should definitely look into. [13:39] if you allow the api to set the hostname of a node, then someone can easily take a DB that has "MAC:preferred-hostname" and populate maas from that. [13:40] allenap: actually, we sort of mentioned that on the call this morning: having "two" hostnames: the ec2 like one and the manual one. [13:40] But we thought about that on the DB level. As smoser said, CNAME is exactly that at the DNS level. [13:41] Yeah. We also mooted allowing nodes to specify their preferred hostname in the DHCP request and making that a CNAME of the "ugly" name. [13:42] But that doesn't cope with collisions well. [13:42] well, its easily fixable anyway. [13:42] the only place that that has value is when maas is not running the dhcp/dns server [13:43] and it assumes that something gave this system a relevant IP address based on its mac address. [13:43] er.. relavant hostname based on mac [13:43] and whatever that something is can then be just dumped/queried. and a tool then update the node once its in maas. [13:44] allenap: collisions are not really a problem if we can write the reverse zone files using only ec2 hostnames. [13:44] collisions are an issue if maas isn't serving the dhcp [13:44] allenap: for each zone, we compute the superset classful network and we write the full reverse zone file for that with ec2-like names. [13:44] rvba: I mean collisions of the "nice" names if done by dhcp; two nodes could ask for the same thing. [13:45] Right, that's another problem. I was thinking about the clasless network collision. [13:45] Maybe the term "collision" is not really proper for that. [13:45] I think my brain has had enough of that stuff for this week ;) [13:59] smoser: that's a great idea I think. Maybe there is something I'm missing but it definitely looks like the missing piece of the puzzle. [14:23] allenap: howdy :) so I was looking at your recen changes and was wondeirng, wher's that maas-provision binary? [14:25] smoser: I tried to summarize why I think your idea is great in the email I just sent. We would have the full forward/reverse DNS config written using ec2-like names. Then on top of that we could use nsupdate to add CNAME record to accommodate custom hostnames. [14:25] records* [14:25] rvba, thanks. [14:25] we expose "friendly-hostname" setting in maas via api [14:25] right? [14:26] Yes we do. [14:26] so my 'mac-addr:friendly-hostname' population script is easy. [14:27] (ie, it takes a file full of lines with 'mac-addr friendly-hostname' and just iterates over it calling maas and setting the hostname for each node based on mac [14:27] ) [14:27] roaksoax: It's a front-end for several commands. It will probably evolve into the front-end for all of MAAS (so that we no longer expose twistd or django-admin). [14:27] smoser: yes, that would work. [14:28] allenap: right, but where is it? do we need a wrapper for it (like we did with maas binary)? (see debian/extras/maas in packaging branch) [14:28] smoser: hum, the key we use in the api is the system_id. [14:28] rvba: that is indeed a good idea as I was mentioning to smoser [14:28] but that means that servers wont have ...domain.com [14:28] which might come handy [14:29] when you do inventory [14:29] that's why I think we should consult this with sysadmins [14:29] Well, we might provide a way to set the custom hostname based on tags. [14:29] So that would be possible. [14:30] rvba, somehow i can surely get a node by its mac-id, no? [14:30] Let me check. [14:30] err.. get a system_id by its mac_id. [14:30] rvba: right, there's various possiblities to approach the issues yes [14:30] we need to consider invetory from now on [14:30] and we need to consider Administrators, and Datacenters and the standards they use [14:30] and provide a sokution based on that [14:31] we can't simply impose them something we believe is better when they already follow standards [14:31] smoser: no, system_id is the key really. But that's something we could add very easily. [14:33] rvba, well i think its essential, really. [14:33] smoser: I completely agree. [14:33] smoser: please file a bug. [14:34] smoser: that's something really easy to add. [14:34] The mac is the information that the outside world knows so it makes sense. [14:35] allenap: btw is that part of the maas cli? [14:37] roaksoax: It is, and it will need a wrapper. It's very simple though: python -m provisioningserver [14:39] allenap: right, and what is the maas-cli stuff [14:39] allenap: err apiclient [14:39] roaksoax: Don't know :-/ [14:43] ok :) [16:17] rvba: ping [16:17] rvba: do you how how could you effectively run yui tests? [16:20] roaksoax: the yui tests are run as part of the test suite but you can run only the yui tests. [16:20] rvba: right, I'm looking towards running them in the actual yui3 source package from the archives [16:20] rvba: so I was wondering how is it that you run them? [16:21] and get the ooutput of them as? [16:21] In MAAS, we use SST to do that (see src/maasserver/tests/test_js.py). [16:30] rvba, bug 1027154 [16:30] Launchpad bug 1027154 in maas (Ubuntu) "need way to get system based on mac address" [Undecided,New] https://launchpad.net/bugs/1027154 [16:31] smoser: ta [16:32] rvba: cool thanks! [20:12] anyone able to +1 this: [20:12] https://code.launchpad.net/~smoser/maas/vdenv-fix-mawk/+merge/116067 [20:12] roaksoax, ^ [20:12] smoser: give me a sec :) [20:16] smoser: done!