[03:50] <axw_> hey folks, we're looking at upgrading some juju 1.25 environments running on MAAS (1.9) to juju 2.x. one thing we need to be able to do is change the agent_name associated with nodes. is that possible through the API? doesn't look like it. any recommendations?
[04:28] <mup> Bug #1697986 changed: API returns object is not iterable to valid API request for server deployment <MAAS:Expired> <https://launchpad.net/bugs/1697986>
[10:33] <engel75> hi all
[10:35] <engel75> just a short question about maas (using it the 1st time), I would like to install 9 physical hosts including LACP (bonding) networking, softwareraid, .... How to create a curtin template used by those 9 nodes?
[12:42] <engel75> this guide is still relevant: http://caribou.kamikamamak.com/2015/06/26/custom-partitioning-with-maas-and-curtin-2/   ?
[15:24] <bryan_att> hi anyone - do you know what preseed command I would use to get MAAS to deploy centos to sdb vs sda (sda is my ssd, sdb is my hdd) ? By default MAAS seems to want to deploy to the 1st disk.
[16:23] <tizmo> hello. is there a way that I can give a node a static route during commissioning before it runs the info gathering scripts?
[16:24] <tizmo> i'm using MAAS 2.2.2.
[16:25] <tizmo> I tested out using the Static Route option under Subnets, but that doesn't seem to apply to commissioning.
[16:32] <blizzow> Let's say I want to deploy an ubuntu server using maas but I want the boot and root partitions to be an mdadm raid1 device? Is that possible?
[17:08] <blizzow> Also, I am trying to deploy the standard centos image to a Dell Server. The installation output ends with "curtin: installation finished" and a few lines before that, it says no error reported.  But maas seems to get stuck in a deploying state for a long time only to finish at a failure.
[17:10] <bryan_att> blizzow: that's what we are doing too, and there are a number of potential issues. I can get it to work (but not completely as desired) with a preseed file that I will paste here.
[17:17] <bryan_att> blizzow: here is the example, this addresses (1) issue which is that MAAS does not support static IP config for Centos. If that's not what you are trying to do, this may not apply. I'll add some more known issues and workarounds / WIP too. https://www.irccloud.com/pastebin/14iY2EkF/
[17:21] <bryan_att> blizzow: issue (2) is that MAAS will try to deploy the OS on the 1st disk found (e.g. sda, which  for me is a 256GB SSD and sdb is a 1TB HDD). This can cause various issues, e.g. if Ubuntu was installed on sdb previously the server may boot to Ubuntu, blocking completion of cloud-init (per the MAAS server's visibility of it). I have not yet found a way to force MAAS to deploy on the 2nd disk.
[17:24] <blizzow> bryan_att: I'm currently running dd if=/dev/zero of=/dev/sda bs=1M on the server to wipe that first disk that had an old ubuntu installation on it. I guess you're saying I should append these  adduser commands and network configuration lines to my /etc/maas/preseeds/curtin_userdata_centos file?
[17:24] <blizzow> I'm fine with actually having the machine come up with a DHCP address and reconfiguring it after. I'm only installing centOS to run a firmware update and then blow the machine away with ubuntu again.
[17:25] <blizzow> dd on a 4TB disk takes tooo long.
[17:26] <bryan_att> blizzow: issue (3) is that at least 5 mins may be required for a deploy to complete if MAAS (for deploying Centos) is not set for DHCP over the 1st NIC to come up. Note this is not an issue when deploying Ubuntu. I use  preseed file above which takes effect but only after MAAS ignores DHCP requests for 5 mins from the 1st NIC (we don't have that one connected to a MAAS rack conrtoller yet)
[17:27] <bryan_att> blizzow: you should also be able to release the machine and have MAAS do the wiping for you. I select wipe disk with fast option.
[17:29] <bryan_att> blizzow: re the actual problem you have I'm not sure. the issues I have found are related to (1) using a 2nd disk for install; (2) having multiple NICs and trying to PXE over one other than the 1st.
[17:34] <mup> Bug #1710681 opened: [2.3] Cannot delete an Ubuntu image, "Update Selection" is disabled <MAAS:Triaged> <https://launchpad.net/bugs/1710681>
[18:12] <bryan_att> roaksoax: can you give me an example of a preseed file that will instruct MAAS to deploy on device sdb? I'm sort of stumbling in the direction of this but the docs are very slim and the curtin command line does not match what is supported in the preseed file afaict.
[18:21] <blizzow> My maas controller is has discovery enabled but none of the IPs are getting resolved in the device discovery section.  Is there something I have to do in order to get device discovery to resolve names?  I do a dig -x ipaddresshere from my maas machine and it gives the FQDN of the host at the ip address.
[20:02] <mup> Bug #1710241 changed: Cannot deploy customized ubuntu cloud image <MAAS:Invalid> <https://launchpad.net/bugs/1710241>
[21:07] <roaksoax> engel75: maas supports software raid and bonding . Software raid for Ubuntu only. Bonding for centos as well (but only for 2.3+)
[21:08] <roaksoax> bryan_att: i'll need to look into it, as we dont really support custom storage config for centos. Curtin picks the first disk it finds
[21:08] <roaksoax> blizzow: check the console, it may not have booted onto the disk afterwards
[21:21] <bryan_att> roaksoax: ok, but I think that does not reflect the real diversity of HW e.g. servers with an SSD as sda and HDD as sdb
[21:27] <blizzow> roaksoax: I'm using serial over lan so I'm kind of blind once pxe boot hits. Deploying the default CentOS7 image from maas 2.2 just hangs at deploying. From what I can tell, it forces centos to install to sda but ubuntu I can set to install to sdc which ends up being the boot drive because sda/sdb are drives connected to a pci slot.
[22:46] <roaksoax> blizzow: probably becuase you are not passing kernel params to enable the console
[22:46] <roaksoax> blizzow: but my wild guess is that it tries to boot onto a different disk than the one where it has been installed into
[23:09] <blizzow> roaksoax: any idea what kernel params I should pass to enable the serial over lan console?