2 Jun 2024 |
| @aloisw:kde.org set a profile picture. | 18:01:44 |
| @aloisw:kde.org changed their profile picture. | 18:23:54 |
3 Jun 2024 |
Mic92 | In reply to @linus:schreibt.jetzt also, how can a process in the zombie state still be using CPU?? Is it accounted as userspace CPU usage? | 11:55:46 |
@linus:schreibt.jetzt | can't tell anymore, it's gone now | 11:56:08 |
Mic92 | I think there was a way to move mounts recursivly | 11:56:35 |
Mic92 | Or did I just handcoded that in cntr? | 11:57:00 |
Mic92 | Linux Hackerman: MS_REC? https://github.com/Mic92/cntr/blob/4080e2ae9f04139849840bb10a45ab6aae4c9749/src/mountns.rs#L177 | 11:58:15 |
@linus:schreibt.jetzt | that's for bind-mounting | 11:58:47 |
@linus:schreibt.jetzt | my use case is that I have an installer system and a system to install, and I want to mount everything based on the fstab of the system-to-install | 11:59:08 |
@linus:schreibt.jetzt | (or system-to-repair, as the case may be) | 11:59:14 |
Mic92 | disko-install does this :) | 11:59:31 |
Mic92 | But using the nixos configuration instead of fstab | 11:59:47 |
@linus:schreibt.jetzt | I worked around this specific case using a wonderful awk script to preprocess the fstab | 11:59:54 |
@linus:schreibt.jetzt | but it won't work for overlay mounts, for example | 12:00:05 |
| Tammi (ey/em) left the room. | 14:19:09 |
4 Jun 2024 |
| raitobezarius changed their display name from raitobezarius (DECT: 7248) to raitobezarius. | 11:16:03 |
netpleb | is it at all performant to use LUKs on a network block device (nbd, or maybe there is a userspace way with ublkdev?) over the internet? | 16:06:34 |
netpleb | I was thinking of trying it as a native btrfs backup method but where the remote backup server does not ever have access to the data. It simply exposes a block device that the client can setup with luks+btrfs. | 16:09:45 |
raitobezarius | whether you do LUKS on client side or server side, I'm not sure this is very noticeable | 16:12:59 |
raitobezarius | CloudFlare has an article on the LUKS performance overhead | 16:13:05 |
netpleb | In reply to @raitobezarius:matrix.org whether you do LUKS on client side or server side, I'm not sure this is very noticeable that was my hope/thought as well. The last time I experimented wtih nbd, it worked fine over LAN but the client sort of hung sometimes. So maybe nbd itself is not the right answer? Are there other well established ways of having a remote block device other than nbd? | 16:14:11 |
netpleb | The only other thing I found but am not familiar with is ublkdev. | 16:14:27 |
netpleb | In reply to @raitobezarius:matrix.org whether you do LUKS on client side or server side, I'm not sure this is very noticeable * that was my hope/thought as well. The last time I experimented wtih nbd, it worked fine over LAN but the client hung sometimes. So maybe nbd itself is not the right answer? Are there other well established ways of having a remote block device other than nbd? | 16:14:47 |
raitobezarius | client hanging seems to be a different concern from LUKS | 16:15:46 |
raitobezarius | for remote block device, you can use RDMA I suppose | 16:16:32 |
raitobezarius | but this requires special hardware | 16:16:37 |
netpleb | In reply to @raitobezarius:matrix.org client hanging seems to be a different concern from LUKS thanks. I really appreciate the input. And yes, the client hanging was not at all a luks issue from what I could tell. Was more just the client kernel not being happy if the server disappears. | 16:20:00 |
netpleb | In reply to @raitobezarius:matrix.org but this requires special hardware Unfortunately I only have commodity hardware to work with on this. | 16:20:42 |
raitobezarius | not sure you can realistically have remote block device with the linux kernel without the risk of hanging issues | 16:21:40 |
raitobezarius | and without something like RDMA AFAIK | 16:21:46 |