| fiesh | we're using netplan on our router / server box, its configuration is https://paste.debian.net/1197100/ during maintenance schedules, it seems our ISPs router (123.123.123.1) does a power cycle, leading to a "enp4s0f3 NIC Link is down" / "enp4s0f3 NIC Link is up" cycle. whenever this happens, the default route is dropped and not added back automatically. I can't figure out why this is, what to do | 07:45 |
|---|---|---|
| fiesh | (installing ifplugd doesn't seem like the right approach here, I'd just like the route to remain, and I don't know how it interacts with netplan anyhow). Any hints? | 07:45 |
| slyon | fiesh: with "default route" you mean the gateway4/6 settings? Is there any other networking daemon running in the background (besides netplan/systemd-networkd), like NetworkManager or dhclient? Or maybe the ISPs router sends IPv6 RAs, to reconfigure the default routes... you could try: "accept-ra: false" | 08:05 |
| slyon | mcint: it should apply on reboot. Try "netplan --debug apply" do see what's going on. | 08:06 |
| slyon | Sat0: macvtap is currently not supported, see: https://bugs.launchpad.net/netplan/+bug/1664847 | 08:08 |
| slyon | oh I see you're using "accept-ra: false" already.. | 08:12 |
| fiesh | slyon: yes, the two gateways are dropped. There's no NetworkManager or dhclient. However, since the system hosts many lxc instances, lxd is running. It seems that lxd spawns an instance of dnsmasq -- could that be related? | 09:10 |
| fiesh | doing a `netplan apply` re-adds the gateways correctly after they've been dropped | 09:10 |
| slyon | fiesh: yeah.. sounds like something is changing the default gateway in the background somehow. Is the default gateway removed all together or changed to a different value? Is the static route still there after this change? | 09:14 |
| slyon | Would it be an option to define your default gateway as a static route as well? | 09:15 |
| fiesh | slyon: the default route is just gone | 09:33 |
| fiesh | slyon: that's a good suggestion... I'll give it a shot | 09:34 |
| fiesh | slyon: ok so it work after the `netplan apply`... I'll have to try a link down later when there's less traffic I'll potentially break | 09:37 |
| fiesh | slyon: I am very hopeful... thank you! :) | 09:38 |
| slyon | :) let me know if that works! | 09:38 |
| fiesh | (I'm optimistic since no other routes go missing after the link cycle, so it would seem that adding the default route this way should behave alike) | 09:39 |
| slyon | ack | 09:54 |
| mcint | slyon: thanks, nothing. in lxd *lxc containers, just the 50-cloud-init.yaml file. ok. it says it applied, but doesn't include the new ip address nor nameserver i configured | 12:29 |
| mcint | renderer: networkd | 12:29 |
| slyon | mcint: would you mind sharing your YAML config? | 12:30 |
| slyon | e.g. via paste.ubuntu.com | 12:31 |
| mcint | yeah... will sanitize values | 12:31 |
| mcint | seems likely that it's a container issue | 12:31 |
| mcint | ** (generate:10751): DEBUG: 12:31:03.940: openvswitch: definition eno1 is not for us (backend 1) | 12:32 |
| mcint | ** (generate:10751): DEBUG: 12:31:03.940: NetworkManager: definition eno1 is not for us (backend 1) | 12:32 |
| lukasm | that looks ok, as you said you're using networkd as a backend renderer | 12:33 |
| mcint | https://paste.ubuntu.com/p/K6RBFr277y/ | 12:34 |
| mcint | I had not included one before. also --debug, helped ...remind me i wasn't doing it as root | 12:35 |
| mcint | also seeing: DEBUG:eno1 not found in {} | 12:35 |
| mcint | ...and the device is called eth0@if130. some kind of lxc autogenerated vdev/if | 12:36 |
| mcint | device that exists. I'm clearly missing some fundamental understanding, and I don't think it's even in netplan | 12:36 |
| slyon | mcint: yeah "eno1 not found in {}" is not good. I think you're missing the netmask in your IP address, try "addresses: [172.1.2.3/24]" (or what every your netmask is | 12:37 |
| slyon | but that should throw parsing errors IMO.. | 12:38 |
| slyon | it shows this for me: "test.yaml:5:7: Error in network definition: address '172.1.2.3' is missing /prefixlength" | 12:39 |
| slyon | try fixing this up, maybe it can then correctly generate your configs | 12:39 |
| mcint | oh, sorry - bad hand-redaction. I do have netmasks on every address for the interface | 14:06 |
| mcint | https://pasteboard.co/K1pi6rs.png | 14:14 |
| slyon | mcint: ok. Is this definition for eno1 or eth0? I think you should not add the @if130 suffix to eth0 | 14:16 |
| slyon | also, indentation might be off due to the #eno1 comment, not sure how picky YAML is in this case | 14:16 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!
