| drab | bah, still stuck on this macvlan thing. I've read all the links I could find, including stgraber's comments but still not joy, unclear what I'm doing wrong: https://github.com/lxc/lxd/issues/1343 | 00:58 |
|---|---|---|
| drab | https://lists.linuxcontainers.org/pipermail/lxc-users/2010-July/000557.html | 00:58 |
| drab | that in theory seems pretty clear... | 00:58 |
| drab | basically add a macvlan interface on the host and have the containers use that as parent with a macvlan bridge type | 00:58 |
| drab | makes perfect sense, but then no joy | 00:59 |
| drab | only thing I know that looks fishy is that lxc network show doesn't show mvlan0 | 00:59 |
| drab | even tho my container is using it and it works (can get an ip, ping etc) | 00:59 |
| sarnold | drab: pastebin your /etc/network/interfaces and lxd configs and maybe someone can spot it? | 01:02 |
| stgraber | drab: what are you trying to do? use macvlan for your containers and still allow them to talk to the host? | 01:12 |
| drab | http://dpaste.com/0G594NT | 01:16 |
| drab | stgraber: yes | 01:16 |
| drab | which yuo commented on in that github issue and on the ML | 01:16 |
| drab | and I think I'm doing what you saif, but I can't get it to work | 01:16 |
| drab | the only thing that seems suspicious is the fact that lxc network list does not show mvlan0 | 01:16 |
| drab | only eth0 | 01:16 |
| drab | and if I try to do lxc network create mvlan0 I get an error that it already exists | 01:17 |
| stgraber | drab: your setup looks wrong | 01:17 |
| stgraber | you should have: | 01:17 |
| drab | ok, great | 01:17 |
| stgraber | - eth0: unconfigured (which seems fine) | 01:17 |
| stgraber | - mvlan0: macvlan on top of eth0, configured with your host's IP (which seems fine) | 01:17 |
| stgraber | - containers: macvlan on top of eth0 (that's wrong in your setup) | 01:18 |
| stgraber | so set parent to eth0 for your containers and that "should" fix the problem | 01:18 |
| drab | ah | 01:18 |
| stgraber | macvlan devices can never talk to their parent, they can only talk to the outside and to their siblings | 01:18 |
| drab | right, which is why I set the parent to mvlan, I was hoping to be able to ping eth0 of the host | 01:19 |
| drab | but that didn't work | 01:19 |
| stgraber | oh, I missed that, your eth0 is also wrong indeed | 01:19 |
| drab | since the eth0 of the host is what is in dns and nodes will try to raech | 01:19 |
| stgraber | your eth0 should be completely unconfigured. What used to be on eth0 should be moved to mvlan0. And then all containers should set parent=eth0 | 01:19 |
| drab | oh, so that's basically the same thing you'd do with a bridge then | 01:20 |
| stgraber | yep | 01:20 |
| drab | I thought the point of mvlan was that I didn't need to touch eth0, but guess that's true only if I don't care about host/guest ping | 01:20 |
| drab | well, point, easier to setup, fewer things to muck with | 01:21 |
| stgraber | right, you don't need to touch eth0 so long as you don't need to talk to it | 01:21 |
| stgraber | which is the case for a lot of people | 01:21 |
| drab | is there still a point in macvlan then? just the supposed better perf from container to container? | 01:21 |
| drab | assuming you wanna talk to the host | 01:21 |
| drab | that is | 01:21 |
| stgraber | there should be a bit more hardware based optimization for macvlan than for standard bridges but bridge+veth is also a much more tested setups so unless you need to squeeze every bit of performance, I'd recommend bridging | 01:26 |
| === Shadowmm_ is now known as shadowmm | ||
| === mwsb is now known as chu | ||
| === madsa is now known as Guest5519 | ||
| === mwsb is now known as chu | ||
| === JanC is now known as Guest79186 | ||
| === JanC_ is now known as JanC | ||
| === andol_ is now known as andol | ||
| === Agent_ is now known as Agent | ||
| spat | Can someone tell me how it is possible to have a disk that is 54% full and complain there is no free space? | 18:15 |
| alkisg | what's the output of df -h, and the exact message of disk full? | 18:17 |
| cncr04s | your inode space could be full | 18:32 |
| tsimonq2 | spat: What FS are you using? | 18:33 |
| spat | ext4 | 18:33 |
| spat | cncr04s: how can I check that? | 18:33 |
| cncr04s | df -i ? | 18:34 |
| cncr04s | yep | 18:34 |
| spat | df -i | 18:34 |
| spat | Filesystem Inodes IUsed IFree IUse% Mounted on | 18:34 |
| spat | devtmpfs 505187 378 504809 1% /dev | 18:34 |
| spat | tmpfs 505610 1 505609 1% /dev/shm | 18:34 |
| spat | tmpfs 505610 499 505111 1% /run | 18:34 |
| spat | tmpfs 505610 3 505607 1% /run/lock | 18:34 |
| spat | tmpfs 505610 16 505594 1% /sys/fs/cgroup | 18:34 |
| spat | tmpfs 505615 4 505611 1% /run/user/5701 | 18:34 |
| spat | sorry should have pastebinned that | 18:35 |
| cncr04s | you only appear to have tmpfs and no actual / | 18:35 |
| spat | /dev/root 655360 655172 188 100% / | 18:35 |
| cncr04s | ah, yep thats all i needed | 18:35 |
| cncr04s | your at 100% | 18:35 |
| spat | didn´t print because of the / | 18:36 |
| spat | yep noticed that as well | 18:36 |
| spat | how can I see where they are used? | 18:36 |
| cncr04s | your using 655,000 files somewhere | 18:36 |
| cncr04s | du -hc / | 18:39 |
| genii | Almost always it's /var/log or /var/spool | 18:49 |
| tpw_rules | on my 16.04LTS server openvpn doesn't start correctly on boot. its service is enabled with my config and it starts great when i do systemctl start after boot has finished. here are the logs: http://pastebin.com/tDHKnUS7 top is after boot, bottom is after running "systemctl start openvpn@server-tcp" | 19:33 |
| tpw_rules | /join #openvpn | 19:53 |
| patdk-lp | did you, systemctl enable openvpn, and systemctl enable openvpn@server-tcp.service, assuming your config is called for openvpn /etc/openvpn/server-tcp.ovpn or /etc/openvpn/server-tcp.conf | 20:02 |
| tpw_rules | patdk-lp: yes | 20:03 |
| tpw_rules | which is how i get the first part of the log i pasted | 20:03 |
| === ewook_ is now known as ewook | ||
| drab | stgraber: can you comment on the nested container issue and zfs? am I understanding correctly that the combination won't work? | 21:00 |
| drab | https://github.com/lxc/lxd/blob/master/doc/storage-backends.md | 21:00 |
| drab | that page suggests that nesting is not supported with ZFS backend | 21:01 |
| drab | and I need directory storage or btrfs | 21:01 |
| drab | but maybe I'm missing something | 21:01 |
| === Edward123 is now known as EdwardIII | ||
| === EmilienM_ is now known as EmilienM | ||
| === mwhudson is now known as Guest34607 | ||
| === Valfor is now known as Guest45243 | ||
| === PaulePan1er is now known as PaulePanter | ||
| === Agent is now known as Guest95782 | ||
| === Guest95782 is now known as Agent | ||
| === trippeh_ is now known as trippeh | ||
| === Pyrrhic is now known as Pyrrthon | ||
| DK2 | what could a soft lock up cpu stuck indicate to? | 23:02 |
| DK2 | im getting kernelpanics with this error | 23:03 |
| ikonia | hardware error | 23:05 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!