| 6 Jun 2023 |
 oddlama | In reply to @elvishjerricco:matrix.org
meaning with the current implementation, we won't have a way to do flushBeforeStage2 = false What would be the benefit of not flushing? | 15:48:17 |
| oddlama | In reply to @elvishjerricco:matrix.org
oddlama: Can you try this in your config and see if it fixes your issue?
boot.initrd.network.flushBeforeStage2 = false;
boot.initrd.systemd = {
services.systemd-networkd = {
before = ["initrd-switch-root.target"];
conflicts = ["initrd-switch-root.target"];
};
sockets.systemd-networkd = {
before = ["initrd-switch-root.target"];
conflicts = ["initrd-switch-root.target"];
};
};
Yep this works :) | 15:50:16 |
| oddlama | Do you have an idea why it doesn't happen on your system? | 15:52:42 |
 ElvishJerricco | In reply to @oddlama:matrix.org
What would be the benefit of not flushing? If your rootfs is on a network fs for instance, you need to avoid flushing. | 15:53:17 |
 @janne.hess:helsinki-systems.de | In reply to @oddlama:matrix.org
What would be the benefit of not flushing? Network mounts | 15:53:27 |
| @janne.hess:helsinki-systems.de | Ah yes :D | 15:53:34 |
| ElvishJerricco | it appears that the person on the systemd issue is wrong that systemd-networkd flushes these things by default. But I'm having trouble parsing through all the documentation and code in systemd on the matter | 15:53:59 |
| ElvishJerricco | If we can get rid of that stupid custom service and just use network config files to control flushBeforeStage2, that would be great | 15:54:26 |
| ElvishJerricco | * If we can get rid of that stupid custom service and just use network config files to implement flushBeforeStage2, that would be great | 15:54:42 |
| ElvishJerricco | In reply to @oddlama:matrix.org
Do you have an idea why it doesn't happen on your system? No idea at all | 15:55:54 |
| oddlama | I frequently think I need a systemd deep dive to grasp any of that properly :D | 15:56:59 |
| ElvishJerricco | Ok I think I have a crude fix for flushBeforeStage2 to use KeepConfiguration instead of our screwy service. It's not meaningfully different, but it's certainly cleaner. Though my code for it could use some cleanup... | 20:59:44 |
| ElvishJerricco | That's a separate thing from oddlama's problem, so I guess I'll make these two separate PRs | 21:00:28 |
| 7 Jun 2023 |
| ElvishJerricco | oddlama: Turns out your bug was hitting my other system; it just didn't have any tangible effect because my network config was the same between stages :P | 20:34:18 |
| oddlama | Ah, that makes sense :D | 20:56:09 |
| 11 Jun 2023 |
| ElvishJerricco | https://github.com/NixOS/nixpkgs/pull/237154 | 06:36:56 |
 K900 | Wait so does this mean we can actually get from stage1 to stage2 without dropping the network? | 06:37:50 |
| ElvishJerricco | K900: That's always been possible with flushBeforeStage2 = false | 06:38:05 |
| ElvishJerricco | This just implements that in a sane way | 06:38:11 |
| K900 | Oh | 06:38:13 |
| K900 | I haven't really tried it because the only box I'd want to try it on gets confused when the network drops while it's booted | 06:38:41 |
| K900 | (for some reason) | 06:38:42 |
| ElvishJerricco | in fact the non-sane thing we did before was the implementation of flushBeforeStage2 = true, because systemd-networkd expects you to want to carry network configs over between stages | 06:39:06 |
| ElvishJerricco | * in fact the non-sane thing we did before was the implementation of flushBeforeStage2 = true, because systemd-networkd expects you to want to carry network configs over between stages and just does that by default | 06:39:20 |
| ElvishJerricco | now the real question is... do I upstream my tailscaled in initrd implementation... | 06:40:30 |
| K900 | Does it need special setup? | 06:41:24 |
| ElvishJerricco | yea | 06:41:30 |
| ElvishJerricco | where is the tailscale state dir? | 06:41:37 |
| ElvishJerricco | because if it isn't something persistent then it's useless | 06:41:47 |
| K900 | Oh yeah makes sense | 06:41:54 |
