[04:23] Bug #1523091 changed: dhcp/dns inconsistency prevents enlistment [04:29] Bug #1523091 opened: dhcp/dns inconsistency prevents enlistment [04:32] Bug #1523091 changed: dhcp/dns inconsistency prevents enlistment === cpaelzer_afk is now known as cpaelzer_ === rbasak_ is now known as rbasak === cpaelzer is now known as cpaelzer_afk === cpaelzer_afk is now known as cpaelzer [11:39] Bug #1532271 changed: maas fails to deploy ibm power nodes === roaksoax-brb is now known as roaksoax [11:42] Bug #1532271 opened: maas fails to deploy ibm power nodes [11:57] Bug #1532271 changed: maas fails to deploy ibm power nodes [11:57] Bug #1542287 opened: BMC fails after node is commissioned === rvba` is now known as rvba === cpaelzer is now known as cpaelzer_afk [12:03] Bug #1542287 changed: BMC fails after node is commissioned [12:09] Bug #1542287 opened: BMC fails after node is commissioned === cpaelzer_afk is now known as cpaelzer [13:48] Bug #1542324 opened: Nodes enlist without any IPMI/BMC information [13:48] Bug #1542326 opened: Cannot deploy Ubuntu 14.04/15.10 on EFI system after 1.8 version to 1.9 upgrade === cpaelzer is now known as cpaelzer_afk === cpaelzer_afk is now known as cpaelzer [14:58] Bug #1542349 opened: Network aliases are created during commissioning [14:58] Bug #1542353 opened: 6 Core system is listed as having a single CPU [15:11] I'm playing around with a new MAAS setup (for testing) and there are a few things I'm stuck on: [15:13] 1. Network topology. I don't understand the intended design. I have a public LAN that is connected to the internet and uses real IPs (134.X.X.X), and my maas region+cluster controller has one interface in this LAN and one interface currently not connected to anything [15:14] I understand from documentation that it's typical to have a second interface on the cluster controller plugged into an extra LAN on an extra switch for this cluster, which is where the nodes will be connected. The cluster controller's DHCP manages this LAN [15:14] But then how are nodes expected to connect with the internet? [15:15] And am I supposed to use private IPs here or another public (global) IP block for this cluster? [15:15] If the former, does that mean the cluster controller needs NAT? How do I make the nodes reachable, then? [15:15] Or is the intended design that *every* node has two NICs, one connected to the internet via the public LAN and one connected to the local network for DHCP/pxeboot/management purposes? [15:16] But that would require having two NICs on every single machine I want to manage, which seems excessive [15:17] Alternatively, I could consider having a “flat” network setup where I reserve a real IP range on the (non-maas) DHCP server running on the “public”/internet-connected LAN and manually configure this with fixed addresses, then set it to use next-server for pxe-boot purposes [15:18] But then I wouldn't be using the maas DHCP at all, and I understand this comes with configuration annoyances beyond the overhead of setting up the DHCP once [15:20] 2. Node commissioning. For now, I've just created a few VMs for testing on the “public” LAN and set the external DHCP server up to load pxelinux.0 and set next-server as the IP of the maas region/cluster controller as described in my previous message. When adding the nodes in the maas web GUI, it shuts them on, boots into its own image, and a minute later or so the status changes to “Failed [15:20] commissioning” with no further debug output [15:21] I watched the terminal output of the VMs while it was commissioning and all the messages go by really quickly, but I do see a lot of apt failures referring to being unable to connect to external domains. I assumed it was some DNS/DHCP error, and decided to boot a livecd for testing purposes (ubuntu), but this livecd has no problems reaching ubuntu.com [15:21] So I'm not really sure why the commission image seems to be silently failing with no logs that give me any information [15:29] https://insights.ubuntu.com/2015/04/10/maas-network-layouts-for-the-landscape-autopilot/ seems to go into some detail about network topology but it leaves me with more questions than answers [15:29] 1. in the “flat network” approach, why does one of the devices still require two NICs plugged into the *same* network? That just seems excessive! [15:29] 2. In the “split network” approach, why are both networks plugged into the same switch? Huh? [15:30] Is the switch supposed to have VLANs configured for them? Why not just draw two switches then? [15:30] And it still doesn't explain how the nodes are supposed to be able to reach the internet [15:30] If they're stuck on a private network [15:41] This graphic would be much more helpful with real example IPs [15:43] https://www.mirantis.com/blog/understanding-options-deployment-topologies-high-availability-ha-openstack/ is not maas-specific but it suggests https://www.mirantis.com/wp-content/uploads/2012/09/netlayout.png [15:44] So I guess that answers what I'm confused about: Yes, you really are supposed to get two NICs on every single node, one for public and one for private usage [15:47] Now I just need maas to actually work :p [15:59] haasn: if you are running on single machine with region/cluster on it, you can have maas on a flat network. You may need 2 nic's connected to the cluster if you want to power manage machine and these are in a different netwpork, if they are routed, then you dont [16:01] haasn: however, in reality, if you only had 1 NIC, power managed is routed, and you have a gateway (aka your router), then you only need to configure the Cluster Controller interfac efor which you want to manage DNS/DHCP and that's were MAAS will pxe boot from [16:03] I think I've decided to stick with a flat network approach and just disable the maas DHCP server, that way I only need one NIC per node and the only drawback is that I have to manually add all new nodes. I don't expect to have more than 10 nodes at most [16:03] So this is fine [16:03] I also don't need an extra switch or upstream configuration [16:07] To clarify about my commissioning troubles: The machines are recognized, I get a full YAML output of all details including CPU, memory, disks etc. which are all recognized successfully [16:08] So clearly it's communicating right, the machines are recognized successfully.. and then it just sets the status to “Commissioning failed” [16:08] With no further debug logs [16:09] haasn: did you see to commissioning scripts section at the bottom and see whether anything failed? [16:10] haasn: so if you go to the details page, at the bottom you'll have lshw output, and a drop down to chose between commissioning output and the scripts [16:10] and maybe one of the scripts is failing, which is causing commissioning to fail [16:10] that's if you are using 1.8+ though [16:16] roaksoax: ah, I missed that dropdown menu [16:16] I get a lot of failed scripts, e.g. https://0x0.st/X2V.txt [16:16] 134.60.209.143:8000 is the maas controller [16:16] But nothing is running on port 8000 [16:17] maas-proxy is installed on the machine, but I didn't touch its configs. The documentation doesn't seem to mention anything about maas-proxy or it being required === dimitern_ is now known as dimitern [16:46] Hmm. Seems maas-proxy.conf is configured to use http_port 8000 and http_port transparent 3128, but the former is ignored by squid3? [16:53] I'm not sure if maas-proxy.conf is being used at all.. it isn't mentioned in the command line of the running squid3 instance [17:06] I found the error [17:06] squid3 was already running with the wrong config (why?) [17:07] stopped it, started maas-proxy manually and now it works [17:07] p.s. does this mean I'm running an open proxy on that machine that the entire internet can access? [17:10] Seems like it, “allow localhost src all # TODO: change this” nice [17:10] Also, wake on lan doesn't seem to work. It works if I run it manually (ether-wake -i enp1s1f0 MAC) [17:10] But maas doesn't get that [17:13] apt install wakeonlan # fixed it === marcoceppi_ is now known as marcoceppi [17:25] Bug #1542410 opened: [1.10] IPMI power query fails [17:28] Bug #1542410 changed: [1.10] IPMI power query fails [17:31] Bug #1542410 opened: [1.10] IPMI power query fails === frobware_ is now known as frobware === narindergupta is now known as narindergupta_aw [19:40] Bug #1542410 changed: [1.10] IPMI power query fails [20:35] The MAAS documentation and advertising suggests I can “Setup the RAID and Network configuration you want through the MAAS web UI or CLI” but I can't find any way to setup RAID on the web interface [20:35] CLI-only? [20:36] Or does that require changes to the preseed?