=== frankban|afk is now known as frankban [09:36] Bug #1702438 opened: [2.2] No way to specify protocol when adding a VMware chassis [13:49] http://paste.ubuntu.com/25025233/ [13:50] trying to deploy to a node after clearing the default storage configuration and specifying raid0 across three drives [14:10] I am assuming it is having problems clearing the partitions [14:10] Eventually failing on 14.04 with "/sys/class/block/dm-11 had no syspath (/sys/class/block/dm-11)" [14:19] gimmic: that seems like a curtin issue. Curtin should be clearing what was leftover on the disk before proceeding [14:19] Yeah. I saw a curtin update and hoped it fixed it, not so lucky [14:31] To test, is there a good way to boot the machine to a workable state I can clear the drive configurations out of myself? [14:31] Maybe recommission the host and allow ssh? [14:41] gimmic: rescue mode ? [14:42] Hadn't used rescue mode yet. Commissioning w/ ssh on worked. I verified /dev/sda4 was still allocated to a vg [14:44] oh, unless that was the commission bits. Need caffiene [14:46] gimmic: no, commissioning doesn't do anything with storage, so you are correct [15:36] QUESTION: What port/service does MAAS talk to in adding (enlisting & commissiong) new machine? [15:36] I am seeing the following error [15:36] maasserver.websockets.protocol: [critical] Error on request (88) machine.action: No rack controllers can access the BMC of node: servername [15:37] MAAS and the HW management network (Cloud service) are on a different Network. So, the issue is similar to this https://bugs.launchpad.net/maas/+bug/1547275 [15:37] Now I need to go talk to Cloud service team to open up a port so it can communicate with MAAS. What port/service (http/https/ssh ...) should I have them open? [15:38] tlian: that seems like your rack controller cannot communicate with the BMC's of the machines [15:38] tlian: I'm guessing that's IPMI machines [15:38] so that would be IPMI ports [15:38] yes. correct [15:39] roaksoax: thank. I will give that a shot [15:41] hi! I am having a little issue as well... with commissioning [15:41] are there some extra settings for the using MaaS behind a http_proxy, which are not just adding the proxy address on the controller? [15:43] While attempting to commission a node, it ends up with "Failed commissioning" and the syslog says that the systemd-timesynctl was timed out [15:47] ... and why is the node not getting the http_proxy variable as global? ... yes, apt works, but the rest of snapd and stuff keep producing errors [15:47] roaksoax? [16:10] Bug #1702509 opened: [2.2.1] DNS locks up regularly [16:29] roaksoax: yup, if I clean up the LVM the system deploys! [16:29] Is there any way I can script the partition management better? I have hundreds of nodes [16:29] the whole point of maas is that I don't have to worry about touching the bare metal as much [16:29] going to validate it now with a fresh node and see if pre-emptively nuking the straggler LVM fixes it [16:31] Basically, the templated storage pre-config is not good for my environment at all [16:40] Another question.. why doesn't maas show me what the dhcp leased address is for a node? It knows what it is, it knows the dhcp pool, it knows the arp table.. I can manually look it up but (auto assign) doesn't help much. [16:40] It should show (auto assigned: 10.10.20.23) [16:46] Bug #1702517 opened: Postgres installed with MAAS logs very aggressively === frankban is now known as frankban|afk [16:53] roaksoax: so the installation process seems to fail to properly remove the vg 'vg_lscratch' during deployment [16:53] I wish there was a "erase disk partitions prior to deployment" checkbox [16:54] seems like that would clear it all up [16:54] maybe during comissioning [17:00] gimmic: 'release' your failed deployment machine and erase the disks : ) [17:00] gimmic: or you can create your own commssioning script [17:02] roaksoax: I am just looking into that. I just want to try to get the http_proxy variable set as global [17:04] I have tried adding something after driver_04_load in the /etc/maas/preseeds/curtin_userdata but it does not take it while commissioning [17:10] roaksoax: I think even saying erase the disks is failing to clean up this lvm mess [17:14] these nodes are hosed. Easiest way I found to clean up the partitions is just to dd zeroes at the start of each.. ugh [17:55] Bug #1702527 opened: cannot delete already existing subnets === Guest39045 is now known as med_ [18:05] Are there any examples of commissioning scripts? [18:06] roaksoax: telling the failed node to erase disks seems to have hung [18:06] gimmic: I am also looking into that right now [18:07] There is this page: https://insights.ubuntu.com/2017/06/02/customising-maas-installs/ [18:07] I assume my issue is related to the same reason curtin is failing, "WARNING: Duplicate VG name vg_lscratch" and issues removing it cleanly [18:07] but it seems a little outdated [18:08] My issue seems to arise from how we repurposed drives from similar nodes to populate out the systems [18:08] I guess both of our problems could be solved the same way: finding out how to modify the preseed [18:08] so we have duplicate vgs in lvm, but I don't even want to use LVM. I just want to nuke the disk partitions from orbit and install [18:09] gimmic: and it is failing while commissioning, right? [18:09] Initially. Then it fails during deployment too [18:09] I cannot even manage to get the commissioning properly [18:11] a good question would be... on the older documentation, it says that one could modify the preseeds at /etc/maas/preseeds, but at the moment they look quite cryptic [18:11] the "commissioning" file contains just "{{preseed_data}}", but where is that defined?? [18:46] gimmic: still there? [18:46] Yup [18:46] I just found out how it works [18:47] still poking around [18:47] it's actually quite simple [18:47] your thing is more tricky, but you can probably get a lot done with a commissioning script [18:48] do you already know how to do it? [18:48] im currently looking at throwing them all into rescue mode and simply tackling it with a remote bash script [18:49] I would still need a custom deploy script to do some other tweaks, like setting up the storage automagically and allocating matching IP addresses [18:49] but the commissioning script part, you already understand it, right? [20:21] gimmic: it could be that it is actually erasing which takes time [20:21] gimmic: therey should be a quick erase option [20:22] Bug #1702560 opened: faild deploy windows 2012r2, but boot stay ok [21:04] Bug #1702567 opened: Make package-dev does not include maas_api_helper.py [23:28] Bug #1702517 changed: Postgres installed with MAAS logs very aggressively