[03:44] <arooni> so this is my gparted... i want to move the unallocated space under /dev/sda2 => /dev/sda1 (which is the boot and / partition)... but b/c they're on separate partitions i'm stuck.  (they are on the same physical drive though) https://imgur.com/a/Wo5bJ6b
[03:44] <arooni> asked in #ubuntu didnt really get any help
[06:38] <vlt> arooni: What did you ask there?
[07:29] <lordievader> Good morning
[07:30] <lordievader> arooni: You  can... but it will be a long and complex process.
[07:30] <lordievader> First you move sda5 and sda6 to the beginning of sda2.
[07:31] <lordievader> Then you shrink sda2 to the space of sda5 and sda6.
[07:31] <lordievader> Then you move sda2 to the end of the drive (moving the unallocated space between sda1 and sda2).
[07:31] <lordievader> Then you extend sda1.
[07:32] <lordievader> IMO, it ain't worth it. I'd make a backup and reinstall with LVM. With LVM you don't have  these problems.
[11:25] <maswan> hm. the official guide seems to suggest that ubuntu can run on 256MiB VMs, but I have automatic updates failing here on a 512MiB host due to dpkg OOM:ing after eating >200MiB ram. Is this expected (and documentation should be updated), or dpkg/foo-issue?
[11:27] <vlt> maswan: 256 MB plus swap could work.
[11:27] <maswan> ah, if I click through I get 512/1G recommended
[11:29] <maswan> hm. I do have 512M of swap too, but it doesn't seem to use it
[11:30] <maswan> ah
[11:30] <maswan> vm.swappiness = 0
[11:36] <maswan> (so, mostly ignore, I'll work on figuring out why it had a different vm.swappiness setting (from boot?) compared to what I get when I run sysctl -p --system...)
[11:50] <mrtAkdeniz> Hey there
[11:51] <mrtAkdeniz> I'm trying to install xserver org and Nvidia drivers to ubuntu server
[11:51] <mrtAkdeniz> But seems like I'm having trouble
[11:52] <mrtAkdeniz> https://gist.github.com/iquad/f937d8ead854ce206a447dc4896da23d here is the issue
[12:11] <blackflow> apt install patience
[14:54] <smoser> o/
[14:54] <rbasak> o/
[14:54] <Ussat> blackflow, I need to install that on my Director :)
[14:55] <rbasak> smoser so the bits I care more about right now are more unit-type tests for git ubuntu build's changelogify and quiltify type functionality (that's horribly broken right now)
[14:55] <smoser> i hadnt looked at/seen the source_builder
[14:56] <rbasak> I'm keen not to block actual progress.
[14:56] <smoser> but i'm fien to try to add tests that way. i agree we need more coverage there.
[14:56] <rbasak> It'll probably need some refactoring to make things more testable.
[14:57] <smoser> we are still in the scenario where we have to very carefully gate entry to trunk/edge ?
[14:57] <rbasak> And I'm happy for that to happen provided that it's being done directly for testability
[14:57] <rbasak> For "build"? I think so.
[14:57] <smoser> rather than just being reasonably careful with trunk/edge and allowing testing there before stable
[14:57] <smoser> ie,m why do we have 4 channels if they have all the same promises
[14:57] <rbasak> Perhaps I should be more revert-happy
[14:58] <rbasak> Especially for build
[15:08] <FastZ> I built a small file server for home use some time back using Ubuntu Server and at the time, only put in a single HDD. I will be cannibalizing another of my computers and want to move the HDD from that one to the Ubuntu server (hdd is same make/model/size) and set up a RAID array to add some redundancy there.
[15:09] <FastZ> Is this possible to do without having to reinstall Ubuntu on the server or must I configure RAID on the two disks during a reinstallation?
[15:10] <FastZ> the server is running on an old Dell Optiplex 380 fwiw, so no RAID controller to work with.
[15:13] <sdeziel> FastZ: that's probably doable but would be an interesting challenge
[15:13] <sdeziel> FastZ: in other words, I'd try only if you: 1) have the time, 2) have the interest to learn it and most importantly 3) have known good/tested backups
[15:14] <FastZ> yeah, i wonder if it would be worth the trouble honestly.
[15:15] <FastZ> to try and set it up without reinstalling the OS that is.
[15:18] <jlacroix> Hello everyone. I've Googled for this and can't seem to find a solution. (Maybe my search skills suck.) I have an older Poweredge server with an iDrac card. The web console of the iDrac card works great. The console requires Java, and I can't seem to get that to work. Is there something I'm missing? I think most browsers don't seem to support Java anymore.
[15:19] <FastZ> there's a setting on the main page of the idra web interface that you should be able to use to configure the console to use HTML5.
[15:20] <FastZ> On the idrac page, to the right of the console thumbnail, the settings link I believe it is
[15:20] <jlacroix> Thanks FastZ, I will try that. It's a Powerdge T410 if that matters, from 2009 or so, I believe
[15:21] <FastZ> oh, i can only vouche that this option is present for idrac6 or 7 and above. I think the oldest idracs we have at work at 6, but can't recall if they have the HTML5 option there or not. I know 7 and up do though.
[15:22] <jlacroix> Thats good to know, I will check that when I'm back home
[15:23] <blackflow> Ussat: heh :)
[15:33] <hateball> jlacroix: If you use Chrome there is https://chrome.google.com/webstore/detail/cheerpj-applet-runner/bbmolahhldcbngedljfadjlognfaaein
[15:33] <hateball> jlacroix: I've used it with... varying results
[15:33] <jlacroix> Thanks hateball
[16:02] <teward> rbasak: if you're around, do you have the ability to set a bug series approval on a bug for me?
[16:02] <teward> ran into a partclone bug that is pretty nasty in Xenial environments when working with large disk imagews.
[16:02] <teward> and the bug needs a series approval for Xenial
[16:02] <rbasak> Yep
[16:09] <rbasak> teward: did you have a bug number for me? :)
[16:13] <teward> rbasak: disregard, hggdh is helping in -bugs
[16:13] <hggdh> done
[16:15] <teward> hggdh: thanks
[16:39] <rkantos> How should I be able to omit the default route when using dhcp-client? I'm having the same issues as here https://unix.stackexchange.com/questions/397368/prevent-eth0-from-adding-default-route
[16:46] <blackflow> rkantos: via /etc/dhcp/dhclient.conf I reckon
[16:46] <blackflow> ah yes, as the first answer in that SO post suggested.
[16:47] <teward> cyphermox: is there a way with netplan to get a DHCP IP address but override other bits like DNS server so that DHCP doesn't set them?
[16:47] <teward> is it as simple as setting those bits while leaving dhcp4 enabled to true?
[16:47] <rkantos> teward: I asked about this @ #netplan, and also tried it.. no success
[16:48] <rkantos> I think the dhcp4: yes still makes the dhclient work based on it's config, and thus just do the default stuff
[16:49] <openfire> There's a reason why I just use systemd.
[16:50] <openfire> With networkd, it's trivial.
[16:51] <blackflow> indeedy.
[16:51] <rkantos> openfire: What do you mean? I should make my own network startup script? :S
[16:52] <openfire> What? All netplan does is create config files for systemd-networkd.
[16:53] <openfire> rkantos: You could easily copy the runtime files from /run/systemd/network to /etc/systemd/network, remove /etc/netplan/*, and customize the .network file corresponding to the interface you want to edit.
[16:53] <openfire> networkd is quite simple to configure, I'll gladly help.
[16:54] <rkantos> I think the issue is not with networkd though, but isc-dhclient
[16:54] <rkantos> isc-dhcp-client rather / dhclient *
[16:55] <cyphermox> teward: not yet, but there's code in master to do that
[16:55] <openfire> rkantos: I'm saying that you're going about this the hard way, if your goal is to ignore a DHCP gateway.
[16:55] <rkantos> openfire: oh you mean I should just delete the route after dhclient runs??
[16:55] <teward> cyphermox: ack.  thought that might be the case.  i have individual overrides in my DHCP for servers that don't need to inherit the rest of the DHCP scope configs, but thought I'd ask anyways.
[16:55] <openfire> rkantos: ... No.
[16:56] <cyphermox> rkantos: what does your dhcp is networkd.
[16:56] <openfire> rkantos: You're using netplan, right?
[16:56] <teward> cyphermox: thanks for the info :)
[16:56] <rkantos> openfire: yes
[16:56] <teward> rkantos: (cyphermox is netplan god by the way... it's why I pinged them directly with my question heh)
[16:56] <openfire> rkantos: That means you're using networkd. netplan CANNOT generate the correct configuration to ignore a route, it does not have this capability.
[16:56] <cyphermox> ^ that
[16:57] <openfire> Along with IPv6 tokens, privacy extensions, etc.
[16:57] <cyphermox> at least not yet, as I pointed out, that's in master, but not released yet
[16:57] <openfire> The EASIEST solution is to skip the netplan middleman, directly configure systemd.
[16:57] <openfire> systemd-networkd, specifically.
[16:57] <openfire> It's absolutely trivial to do this if you're using networkd directly.
[16:57] <openfire> Two lines.
[16:58] <rkantos> openfire: yeah, no I joined #netplan too..
[16:59] <openfire> Okay?
[16:59] <rkantos> But I wasn't using netplan to configure the dhcp server, since the only thing I want to achieve is sending the hostname to the dhcp-server..
[17:00] <rkantos> So netplan wasn't causing "my issue"
[17:00] <openfire> 11:39 <rkantos> How should I be able to omit the default route when using dhcp-client? I'm having the same issues as here https://unix.stackexchange.com/questions/397368/prevent-eth0-from-adding-default-route
[17:00] <openfire> That was your original question.
[17:00] <cyphermox> rkantos: the point openfire is trying to make is that you can do this by configuring the client in networkd directly
[17:00] <openfire> Your hostname question is also trivial under networkd.
[17:01] <rkantos> openfire: I thought that it would lead me to a solution since it wasn't using netplan either, though
[17:02] <openfire> What?
[17:02] <rkantos> and in the comment you can see he had the same issue with dhclient, where changes in config didn't change the behaviour
[17:02] <rkantos> (changes in dhclient.conf)
[17:03] <openfire> Because networkd is controlling the interface, yes.
[17:03] <openfire> So... The fix must be made using networkd.
[17:03] <rkantos> But, sure, I am open to networkd suggestions
[17:03] <openfire> I don't really know what I keep having to argue this point...
[17:03] <openfire> s/what/why/
[17:03] <openfire> You're using networkd. Use networkd to resolve your issues. Mucking with random config files is NOT the best option.
[17:04] <rkantos> I believe you.
[17:04] <openfire> Anyway!
[17:04] <openfire> How many interfaces do you have on the system in question?
[17:04] <blackflow> Ideally use networkd directly as it's all configurable in the one .network file for your connection. UseRoutes directive, see systemd.networ
[17:04] <openfire> Exactly.
[17:04] <blackflow> ... damn enter.  systemd.network manpage
[17:04] <openfire> I don't use netplan.
[17:06] <rkantos> openfire: one or two, but of course mainly looking for the solution to the two if version
[17:06] <openfire> rkantos: So, I think the answer I'm after is 'two.'
[17:06] <openfire> rkantos: I'm assuming both are DHCP configured, but ONE should ignore the gateway.
[17:07] <rkantos> openfire: no, the LAN is static
[17:07] <blackflow> (or both but you set routes manually)
[17:07] <openfire> rkantos: Okay, give me a brief synopsis of how both interfaces should be configured.
[17:08] <rkantos> Bottom line is I want my router to just know the hostname of a static client without having to configure it to it on it, but instead just use dhcp to send the hostname to it. I don't want dhcp setting the routes again.
[17:09] <openfire> This, my friends, is called an XY problem. So your issue is that you have a DHCP server on a router, you want it to send the hostname to the client.
[17:09] <openfire> But only the hostname, not a default gateway, for some reason.
[17:10] <openfire> Why not a default gateway?
[17:10] <rkantos> openfire: no, I want the hostname from client to router
[17:10] <openfire> rkantos: For DNS purposes?
[17:10] <rkantos> yes
[17:10] <openfire> Then none of what you're using is right.
[17:10] <rkantos> :D
[17:10] <blackflow> wait... if the lan is static.... why not.... just put that hostname+ip config in the DNS resolver for your LAN.... STATICally....
[17:11] <openfire> DNS and DHCP are separate processes. You'll have to work up some mechanism to have the DHCP service tell the DNS service.
[17:11] <openfire> blackflow: XY problem at its finest.
[17:11] <blackflow> indeed.
[17:11] <blackflow> the ugly cousin of Yak Shave.
[17:12] <rkantos> I want to keep it "simple" and not change stuff on my router when I create a new vm on my proxmox, basically
[17:12] <openfire> And what OS is your router?
[17:12] <rkantos> Tomato
[17:13] <openfire> Then the issue is still solvable on the client.
[17:13] <openfire> Now, why DON'T you want DHCP to give out a default route, exactly?
[17:13] <rkantos> because I don't necessarily want the client to use internet through the lan
[17:14] <openfire> Sooo... Firewall rules?
[17:14] <blackflow> sounds like you need a whole router+dns setup straight under your control, with that existing router being just an upstram node on the WAN/LAN boundary
[17:15] <openfire> Because, in general, one does not control client access ON the client.
[17:15] <openfire> ip route add BYEEEEE
[17:15] <rkantos> blackflow: yeah, sure I thought about that, but really don't see why I'd really like/need it at this point
[17:17] <rkantos> So I want hostnames to work in my lan(s) under the Tomato router without setting up a whole DNS, and not having to config the static ips on the router every time
[17:19] <rkantos> I thought I would just be able to send the hostname with dhclient; Apparrently not that simple ( I understand if it is not a desired result, thus why it is difficult)
[17:19] <blackflow> then just use dhcp for everything.
[17:20] <blackflow> sending DHCP client hostnames up to DHCP servers is part of DHCP and afaik every client implementation can do that.
[17:20] <rkantos> It doesn't achieve not having to configure the router everytime :D
[17:21] <blackflow> with networkd you can do it with SendHostname directive, see systemd.network(5) manpage.   dhclient has send-hostname option but I think we established that dhclient is just a XY redherring here
[17:21] <theGoat> so i want to use dd to clone a drive, is there a flag i can throw at it to only do used space on the drive
[17:22] <blackflow> so allow dhcp to dish out IPs, sethostnames, let it be the LAN's resolver. on specific clients that shouldn't use the default route, you simply configure the default route manually, with everything else the same, dhcp based.
[17:22] <blackflow> s/allow dhcp/allow the router/
[17:23] <blackflow> theGoat: no. block base has no concept of "used space", that's filesystem, one level above.
[17:23] <openfire> theGoat: dd has no concept of filesystems, so it has no concept of used vs empty. You could consider rsync.
[17:24] <rkantos> blackflow: But doesn't that mean I need to configure every vms static ip on the router everytime?
[17:24] <blackflow> rkantos: no. routers/dhcp-servers have the ability to integrate with DNS resolvers
[17:25] <blackflow> so your clients send out the desired hostname and the dhcp assigns the IP for it and tells to the DNS resolver the combination of hostname + ip
[17:25] <blackflow> router applianes are probably working like that by default.
[17:25] <blackflow> *appliances
[17:30] <rkantos> Hmm, ok.. I guess this is a question to Proxmox.. But since Proxmox manages the networkd configuration (in addition to additional Netplan config); Where can I put the SendHostname config other than proxmox's generated config?
[17:33] <blackflow> I don't know what and how Proxmos does anything, sorry. SendHostname is for systemd-networkd, namely the confg option of a .network unit for a connection.
[17:33] <blackflow> *Proxmox
[17:35] <rkantos> yeah, eth0.network reads: Description = Interface eth0 autoconfigured by PVE
[17:37] <rkantos> It is nuked at every reboot, obviously
[17:38] <blackflow> rkantos: does proxmox even allow dhcp for VMs? I found some older forum posts that say Nope.
[17:40] <blackflow> see, thing with virtualization (again, I don't know how/what Proxmox does specifically, but in general and say with qemu-kvm) is this. on the host side you have a bridge and every VM gets a virtual NIC on it. That's just virtualized hardware. then the VM's own operating system must have a dhcp client that deals with its own networking.
[17:40] <rkantos> The options are there in 5.x at least, and they work
[17:40] <blackflow> so then the SendHostname question is for VM guest OS and its config.
[17:41] <rkantos> yeah, i know.. But it won't work with the default eth0.network config, as it gets wiped out by Proxmox at every reboot (for the lxc containers at least, have not tested with kvm, yet)
[17:42] <blackflow> what won't work exactly? VMs have no clue what eth0.network is, on the host side. the only way host does is a bridge.
[17:42] <blackflow> is Proxmox setting up jsut that? a bridge based on eth0?
[17:42] <rkantos> I think with KVMs, it doesn't change the vms networkd config at all, as it does with LXC
[17:43] <rkantos> blackflow: eth0 here is the name of the interface on the vm itself
[17:43] <blackflow> wait wait wait. Is this virtualization or containerization? some people use "virtualization" for _both_ and it is completely wrong.
[17:44] <rkantos> blackflow: LXC, so I guess containerization
[17:44] <blackflow> the difference being that in a VM, you only, ONLY, get virtualized hardware and your "guest" VM OS needs a kernel to deal with it.
[17:44] <rkantos> yeah
[17:44] <blackflow> containers are namespaces and thus share the host's kernel, IPC, networking, etc..... yuge difference.
[17:47] <plm> Hi all
[17:47] <blackflow> unfortunately I have very little experience with containers and network namespaces, so I'm not sure I can help with that specific issue there. afaik individual containers _can_ have their own dhcp clients becasue the network is namespaced -- if set up that way of course, so it could still be the same solution -- individual dhcp clients in containers.
[17:49] <rkantos> blackflow: yeah, I think it works close enough it has no difference, it has it's own MAC etc, vlans work without issues etc.. The problem is just that the networkd configuration is also managed by Proxmox, so I cannot change the default .network -config.. That was the main reason why I was trying to make dhclient "stupid" (send only the hostname)
[17:52] <blackflow> rkantos: at this point I belive this is very proxmox specific issue.
[17:53] <rkantos> blackflow: well yes of course, since apparretly dhclient / netplan won't work the way I thought or wanted. Already making noise on their irc-channel :P
[17:54] <blackflow> my personal dislike for netplan aside, I believe you should disable it in that setup, unless proxmox is using it directly. but note, netplan is JUST config abstractor. the actual management is done by a backend, networkd probably (because the only ohter supported atm is NM)
[17:54] <blackflow> because too many layers can only mean huge mess and you know.... too many cooks spoil the broth.
[17:58] <rkantos> Hmm.. Now that I think about it.. How can I configure additional network configuration a bit like netplan? The Netplan configuration works (and thus coexists) as it creates a route to a 2nd lan I have..
[17:58] <blackflow> netplan is just abstractor. take a look at .network unit it creates, copy it and modify.
[17:59] <rkantos> can I just create a 1-eth0.network config with additional config? netplan seems to have a /run/systemd/network/10-netplan-eth0.network file on the vm
[18:00] <rkantos> Proxmox generated config is at /etc/systemd/network/eth0.network
[18:00] <blackflow> you can name it as you want, but place it under /etc/systemd/nework. netplan does it under /run/.... because it's using systemd generators for run-time creation of networkd unit from the yaml based config.
[18:01] <blackflow> sounds likea  conflict to me, if they're BOTH configuring for the same interface.
[18:01] <blackflow> (cooks, broth...)
[18:02] <rkantos> maybe this is the reason I need to be here.. As that seems like a solution
[18:02] <rkantos> As I said, the netplan config seems to work though, as it creates the one route that I want.. Maybe I can add just the SendHostname in there somewhere too?
[18:05] <rkantos> If I have just [DHCP] and SendHostname=true .. Will that do other default stuff or will I need to set false to all of them too?
[18:07] <blackflow> not sure what you're asking. netplan doesn't have SendHostname directive, that's networkd. If it has another yaml config option for the same effect, I don't know.
[18:15] <rkantos> blackflow: asking about networkd manual configuration... How should I name a file that networkd would read after eth0.network?
[18:16] <blackflow> rkantos: "read after"? systemd units don't work like that based on filenames. you express before and after relationships in the units themselves
[18:17] <blackflow> but yeah they are processed from disk, in lexical order
[18:18] <rkantos> ok, I misunderstood the "lexical order" bit
[18:19] <blackflow> I mean, if two files deal with the same interface, being processed in that order makes a difference. if that's what you were asking.    also same filenames replace each other ,  eg /etc/.../eth0.network replaces stuff under /run/.../eth0.network
[18:19] <blackflow> wait, I _think_ etc has higher priority... lemme check
[18:20] <blackflow> ah yes. /etc over /run over /lib.
[18:21] <rkantos> yeah that's what I meant / thought
[18:22] <rkantos> If I create config similar to netplan at /etc/systemd/network/1-eth0.network , it doesn't work like netplan did at /run/systemd/network/10-netplan-eth0.network
[18:23] <rkantos> e.g. the routes that netplan had added to /run are not being applie from the file i created @ /etc
[18:23] <openfire> Because the one in /etc overrides the one in /run.
[18:23] <openfire> They're not additive.
[18:24] <rkantos> Huh? Didn't I just say that the one in /run (the Netplan generated one) worked? :S
[18:24] <openfire> No. And I'm tired of you not being clear about things, so I'm going to go do other things.
[18:34] <ahasenack> rbasak: still around?
[18:37] <rkantos> Ok, I appreciate the help
[18:39] <ahasenack> rbasak: n/m, sorry for the ping :)
[18:41] <rkantos> Is there a way to only send the hostname if I have this as the networkd config that unchangeable? https://pastebin.com/VeLPbBda
[18:46] <blackflow> rkantos: no. DHCP = none means no dhcp. no dhcp = no hostname sent  via........ dhcp :)
[18:47] <blackflow> rkantos: take the netplan generated .network from /run and stick that under /etc/...  and modify as you want. then disable netplan (remove any configs from /etc/netplan/)
[18:48] <blackflow> rkantos: but _again_  if proxmox is managing the NIC, you shouldn't do additional configs like that. either leave it to proxmox or do it all manually and disable even proxmox from doing that.
[18:54] <rkantos> Ok, though doing anything with Netplan doesn't change the issues with sending the hostname.. Again, this was why I thought I could just use dhclient.conf & dhclient to send only the host name. Oh well I guess I'll see if there is Proxmox stuff where I could disable the DHCP = none default
[18:56] <blackflow> at which point why don't you leave the IP assignment to DHCP?
[19:00] <bane500> hey guys, is it possible to have two bonds each using round robin
[19:01] <bane500> using ifenslave
[19:02] <rkantos> blackflow: I think the main reason is I don't want to run the Tomato router WebUi or SSH so that it can be accessed from the LAN.. I could instead add one physical port to it for management only. That would mean I would have to physically connect to the router everytime I want to add static IPs for the DHCP server.
[19:02] <sdeziel> ahasenack: I've open some merge requests on Debian's salsa and haven't yet receive any ack. Do you know if I should open Debian bugs and link them to those MR?
[19:03] <sdeziel> bane500: as in 4 NICs bundled in 2 bonds?
[19:03] <ahasenack> sdeziel: maybe, yeah. I heard that by default salsa doesn't email the maintainers about new MRs, unless the maintainer explicitly chose to opt-in
[19:03] <sdeziel> ahasenack: ah bummer, thanks!
[19:03] <ahasenack> I had such a case once with krb5
[19:03] <bane500> sdeziel: yeah
[19:04] <sdeziel> bane500: each bond is independent so yes
[19:04] <bane500> sdeziel: ah okay. I'm getting an error right now when i attempt to restart networking services. Only 1 bond is coming up, the other bond is erroring out stating 'waiting for a slave to join bond0'
[19:05] <blackflow> rkantos: OR.... use DHCP normally, have the router dish out IPs and see if it works by adding names received via DHCP, to its DNS resolution pool.
[19:05] <bane500> sdeziel: oh and also 'bond0: option mode: unable to set because the bond device has slavs'
[19:05] <bane500> slaves*
[19:06] <sdeziel> bane500: it's possible that you need to select the mode prior to having slaves joining
[19:07] <sdeziel> bane500: my only setups using bonding are running 16.04 but it shouldn't matter much (except for no ifupdown on 18.04). Anyway, with those I have no problem using different xmit-hash-policy on separated bonds
[19:08] <bane500> sdeziel: is it possible that I can only have 1 primary interface
[19:08] <sdeziel> bane500: that would be active-passive mode, no?
[19:09] <bane500> sdeziel: not sure.
[19:10] <sdeziel> bane500: maybe this /etc/network/interfaces extract will help you: https://paste.ubuntu.com/p/vJCFN9Q2cK/
[19:14] <sdeziel> bane500: I now realize I'm using the same mode on both, sorry for the misdirection
[19:14] <bane500> hmmm
[19:14] <bane500> sdeziel: here is the error i'm receiving: https://pastebin.com/PqAa1VFE
[19:15] <rkantos> blackflow: yes, naturally that will work until I want a vm to use a second interface with a public ip using dhcp(e.g. dynamic public ip), unless I can use networkd to then remove the default route from the lan interface.
[19:15] <blackflow> rkantos: you _can_ do that. it's only a question of _default_ confiugration vs individual configuration for a particular VM
[19:15] <sdeziel> bane500: I just tested and I can flip my bond1 to "balance-rr 0" while keeping the bond0 as "802.3ad 4" so my claim is still valid, phew ;)
[19:15] <bane500> hahaha
[19:16] <sdeziel> bane500: how's your /e/n/interfaces file like?
[19:18] <bane500> sdeziel: just pm'd it do you
[19:18] <rkantos> blackflow: How would I achieve that then when taking the Proxmox forced config in to account (https://pastebin.com/VeLPbBda) ?
[19:19] <blackflow> rkantos: I don't know how to work with Proxmox. that's why I mentioned earlier that this is a proxmox issue, not an ubuntu one.
[19:20] <blackflow> but you _can_ have VMs default to regular DHCP networking and have individual VMs override route or whateve you wanna overide.... then again, are those VMs or containers.... :)
[19:20] <rkantos> blackflow: Yes. I'm just thinking that since Netplan configuration can have an effect (and "coexist") to some level with the forced Proxmox config.. I could still do something with networkd - too.
[19:22] <rkantos> blackflow: I'm using LXCs all the time. As I said previously, I think the KVMs wouldn't have this issue, because their addresses cannot be set from the Proxmox WebUI
[19:22] <blackflow> again....... netplan is just configuration abstractor. it creates a .network unit file. if it manages the same NIC as proxmox that ALSO creates a .network unit file, you have conflict on several levels there.
[19:22] <blackflow> if those are not virtual machines but containers, don't use the name "VM" for them please, you'll just confuse anyone trying to understand your problem.
[19:23] <blackflow> container != VM. LXC != virtualization.
[19:53] <rkantos> blackflow: ok
[20:00] <rkantos> hmm... Maybe this could work! If I have the Netplan generated config modified by adding the "SendHostname under [DHCP]" and Proxmox / lxc forced configuration set to DHCP, no default route is set!
[20:07] <blackflow> rkantos: why tho'. if proxmox is handling it, why have netplan in the game.
[20:07] <blackflow> you need to adjust what proxmox does, or force a manual configuration (when then can be manual .network or netplan if you prefer it)
[20:08] <rkantos> _Without_ the additional eth0 config (either manual or Netplan generated..) and dhcp enabled on the lxc, the default routes are set. I thought networkd would also set the routes when specifying some [DHCP] options. If it really works this way, then I can use Netplan to configure additional routes and just add a /etc/systemd/networkd/1-eth0.network with the [DHCP] SendHostname ! I think this is all
[20:08] <rkantos> that I needed.
[20:09] <blackflow> I have this feeling you've ignored everything said so far and keep pusing your initial conflicting configuration. well, good luck. ;)
[20:13] <rkantos> blackflow: if it works with it, I'm satisfied. Netplan comes by default on the lxc templates that you can use with Proxmox, so I guess it cannot be all bad..
[20:16] <rkantos> It seems now that if I set a "static ip" IP in Netplan / networkd, then just enable the Sendhostname and leave DHCP otherwise enabled (from Proxmox for the lxc); DHCP requests the IP configured and doesn't set default route! This is what I wanted. Too bad it is not possible to set an IP with DHCP from the Proxmox GUI.
[20:16] <blackflow> you have two different networking managers competing to configure one NIC. even if it (coincidentally) works, the fact remains that you have two managers trying to configure one NIC independently.
[20:16] <Ussat> ^^ THAT
[20:16] <Ussat> you need to choose one
[20:18] <blackflow> (and managers is wrong word, netplan doesn' tmanage anything, it creates a .network unit form yaml and then pings systemd-networkd to apply it)
[20:21] <rkantos> I don't see it as a major issue if it works and the config that I would do would be exactly the same anyway. (well apart from Netplan not being able to configure the [DHCP]SendHostname)
[20:21] <blackflow> reboot and you'll see :)
[20:21] <blackflow> (after you've changed proxmox or netplan generated .network files, manually)
[20:22] <rkantos> that's what I am doing when I'm testing it, constantly, along with flushing the routing table (ip route flush table main && reboot)
[20:23] <blackflow> then you made the changes persisten through another .network unit?
[20:38] <rkantos> blackflow: Ussat ok you're right :)
[20:39] <rkantos> I'll just add a 2nd .network @ /etc with the "static ip" and [DHCP] and it works.. It seems to work.