=== jtv1 is now known as jtv === freeflyi1g is now known as freeflying [10:56] I may be in the wrong room here -- if so please suggest the right one! Please could someone tell me which package to install onto Ubuntu 12.04 to get the maas-cli client? [11:10] cbeebie, not 100% sure, but i think in the maas ppa apt-add-repository ppa:maas-maintainers/stable [11:20] melmoth, thank you. I have the package now. There's no man page for maas-cli, but I guess that's a tier two issue :-) [11:21] i have never used it to be honest :) [12:52] hi roaksoax i've seen this many times now and i thought i would report it: https://bugs.launchpad.net/maas/+bug/1155607 [12:52] Launchpad bug 1155607 in MAAS ""DNS zone name" in Cluster controller and "Default domain for new nodes" in settings should be the same" [Undecided,New] [12:52] but i'm not sure if it's on purpose [13:30] racedo: well the thing is that ypu can assign different zone name depending on the cluster [13:30] so that eould defeat the purpose of making them match === kentb-out is now known as kentb [15:28] hello [15:29] i have 5 servers running centos [15:29] i want to install ubuntu server 12.04 [15:32] i need more information about maas [15:34] johangm90, http://maas.ubuntu.com/ [15:34] i m not sure i understand what you are tryint to do exactly, nor how maas is involved. [16:02] roaksoax: you around? [16:04] trevorj: hi [16:05] roaksoax: hey! I heard you are the one working on the IPMI related stuff in maas? [16:05] trevorj: i am :) how can i help you? [16:05] roaksoax: Well, first of all, great job [16:06] roaksoax: second, I've bene writing some vendor specific stuff for HP iLO [16:06] roaksoax: and I'd love to get it included into maas somehow [16:07] roaksoax: pretty much, I want to set the hostname in iLO to the hostname of the actual machine, for easy finding in a big blade enclosure, and do simple bios setup, such as setting PXE boot to primary [16:07] roaksoax: I've gotten it somewhat working in dev, but instead of writing HPONCFG xml files as I used to, I used a library called python-hpilo [16:08] roaksoax: The problem is it to access local hpilo devices, you currently (afaik) need to use the hponcfg binary from HP, which I'm unsure of the distribution licensing on. I believe it may operate over simple serial through the hpilo kmod, however [16:09] trevorj: so setting PXE boot to primary you might not need to. Everytime maas tells a machine to power on, it will tell it to PXE as well [16:09] roaksoax: Oh, really? Is this a new thing? [16:10] trevorj: nope, it's been working like this for a while now [16:10] since the early implmentation of mAAS IPMI power management [16:10] roaksoax: Hmm, I think I might of just been hitting the IPMI bug you already squashed when dealing with which IPMI user slot to use [16:11] trevorj: probably, that should be released soon in archives (or should be available in ppa:maas-maintainers/stable) [16:11] roaksoax: yeah, I just need to upgrade ;) [16:12] roaksoax: well than fantastic. There are still things I'd love to automate however, like enabling VMX support [16:12] trevorj: now, as far as doing some customization to your iLO, you could do it in 2 ways [16:12] trevorj: 1, would be to do it in during the commissioning step [16:12] roaksoax: lots of HP blades come with vmx disabled, unfortunately [16:12] trevorj: or 2. do it everytime right before the node gets started [16:12] roaksoax: yeah, that's where I've been doing all of the setting changes, in commissioning [16:13] roaksoax: I also made a small agent that runs via upstart as a task that sets the hostname to the current hostname if they don't match [16:14] trevorj: yeah. so all the stuff done in maas in done by freeipmi-tools [16:14] trevorj: currently, mostly using bmc-config [16:14] or ipmi-chassis-config [16:14] roaksoax: yeah, I've been going through the commissioning config =) [16:14] roaksoax: I love how it's rather cross vendor [16:15] roaksoax: I didn't even know IPMI had such standardized user/power mgmt [16:15] trevorj: yeah "standarized" [16:15] :) [16:15] roaksoax: heh, yeah [16:15] trevorj: bu yeah if you are doing that already, then you are in the right track [16:16] roaksoax: Would this be wanted in maas? [16:16] trevorj: maybe :) [16:16] roaksoax: ok, good enough for me [16:17] roaksoax: better question, would vendor specific optimizations be wanted? [16:17] roaksoax: or are you guys trying to stay "standardized" IPMI? [16:18] trevorj: verdor specific optimizations are difficult, specially if it involves the use of privative tools. However, it is always worth the shot [16:18] trevorj: because we are trying to"standarize" its usage in maas, but at the same time allowing people to do its own thing, as you did [16:18] roaksoax: if I could create/find a way to do it with only OSS tools, would that be better? [16:18] trevorj: yes that would indeed [16:18] I agree at heart ;) [16:19] I bet it's just a serial protocol too [16:19] trevorj: i would recommend you file a bout in http://launchpad.net/maas and propose your improvements [16:19] or things that you would like to see [16:19] roaksoax: will do [16:19] roaksoax: thanks! [16:19] oh one more thing [16:20] if commissioning runs more than once somehow [16:20] does it change the maas IPMI login's pass and then maas doesn't allow the request to update it as it already exists? [16:20] or does it grab the existing password and set it to that again? [16:21] trevorj: it will change it, but it should update the credentials in MAAS too [16:21] I know I get a traceback if it runs again, due to the node already existing [16:22] but it may update the credentials, I'll have to test it, as I was dealing with other issues as well at that point [16:22] trevorj: right, the thing would be to see if the credentials have been updated [16:22] trevorj: because in the term of ipmi credentials, we simply "update" the node information [16:22] ah, so there's more than one request? [16:22] trevorj: but yeah, re-running commissioning once a node is already commissioned, it is not really a supported feature [16:23] roaksoax: hehe, it was an accident due to the server name in iLO being unknown ;) [16:23] roaksoax: hence my whole process above ^ [16:24] indeed [16:24] trevorj: and yeah we do updates various times [16:24] trevorj: signal "--power-type=ipmi" "--power-parameters=${power_settings}" WORKING "finished [maas-ipmi-autodetect]" [16:26] Ah, I've seen that in there! Makes sense now. I've got much code to go through in maas still, as I was saying before, it's some great Python to read. [16:26] indeed [16:26] my code is probably the ugliest :) [16:27] nah, it just looks like it grew fast [16:27] yeah [16:28] catting a couple files through a script is something I do at times as well, without a better distribution mechanism it makes perfect sense to me ;) [16:29] :) [16:30] ok, I'll raise it on launchpad, look forward to helping in any ways I can [16:30] looking* [16:31] cool :) [16:31] thanks! [17:27] Hey all, anyone around to answer a quick question about preseed? [18:30] druiid: sup [18:31] Hey, so... I was looking around pretty much everywhere and wasn't seeing how/if you can define per-group preseed configurations. Is this possible? Basically I wanted to define software raid for specific nodes and rely on the hardware-raid for others. [18:34] druiid: hmm, good question [18:36] druiid: There may be a way to assign a preseed on a per node or preferably nodegroup object [18:36] druiid: I'm interested in this as well, because I'll eventually have to assign different preseeds based on more than pure architecture [18:37] druiid: The devs are in here, I'm sure they'll be able to shine more light on if this is currently possible using the existing models [18:38] Hopefully so! Kind of critical for me as there are specific systems that the onboard raid isn't supported very well by Linux (grr, Intel) so software raid is necessary. [18:39] personally I'm a bigger fan of software raid than hardware, but I can see your reasoning [18:39] s/but/so/g [18:40] Well, I also couldn't just use a default preseed even across the cluster and just give them all software raid as some would require raid-6 and some raid-1. [18:40] Is a preseed just a shell script? [18:41] I've always meant to look into that, I have a feeling it may be [18:41] Sorta. More like a configuration file. It's like the Ubuntu version of kickstart (for redhat systems) [18:41] Yeah I've used preseeds before [18:41] but I'm wondering if they are actually ran through sh [18:41] d-i is an installer command [18:42] Oh. Not sure. Probably shell based, yeah. [18:42] If it's actually ran and not just parsed, then you have the benefit of if statements [18:43] actually [18:43] maas uses templated preseeds [18:43] iirc [18:43] that may be the 'proper' way to handle this [18:43] maybe it's generated from the templates per instance [18:44] if so I bet you have variables and if statements etc via {{ }} [18:46] Hmm, that very well could be. There are indeed lots of if statements and using xpath style variables. [18:46] I wonder if the tags get expressed as well. [18:47] I forget the template syntax they use, I'd have to go through that part of the code again, I've only read through certain parts in lieu of issues and skimmed through others [18:48] iirc it's not jinja or mako [18:49] http://pythonpaste.org/tempita/ [18:49] aha! [18:50] there we go ;) [18:51] If they generate it per instance, which would make sense to me, then I bet you have a node object where you can access it's xml hardware info [18:52] Sounds reasonable. A bit clunkier than I'd like, but should be a solution barring another way it's meant to be done. [18:55] I guess that at least gives me a direction for next week :)