| === himcesjf_ is now known as him-cesjf | ||
| alkisg | Hi, can anyone imagine why the netconsole module works properly early in the boot process, yet it stops broadcasting when systemd services start, at about the point when systemd-journald starts? (maybe that's when networking starts too...) | 10:04 |
|---|---|---|
| alkisg | UDP traffic isn't blocked, e.g. sending messages with `nc` works fine; it seems like the netconsole module just stops sending; it even loads successfully with no errors, yet it sends nothing | 10:05 |
| alkisg | Example commands to test with: server: nc -l -u 6666; client: modprobe netconsole netconsole=@client-ip/enp0s3,@server-ip/ | 10:06 |
| alkisg | These work at e.g. init=/bin/bash, but not on "recovery" or later | 10:06 |
| alkisg | I tried downloading netconsole-setup from 19.04; this manages to send a single test message, and then nothing more | 10:32 |
| alkisg | It works in Ubuntu 4.10, fails on Ubuntu 8.04. Hehe, first time trying Warty. | 10:47 |
| jeremy31 | alkisg: Why mess with unsupported versions? | 10:48 |
| alkisg | jeremy31: to see when it broke, to help in finding why it broke | 10:49 |
| alkisg | (netconsole isn't working in ubuntu since at least 10 years now) | 10:49 |
| alkisg | (while it still works fine in buster and all the rest distros) | 10:49 |
| alkisg | 6.06.1 is broken too, so it broke sometime between 4.10 and 6.06.1 | 10:54 |
| alkisg | So, it works on 4.10 and fails on 5.04 and up to now. Ancient bug. :) | 11:05 |
| alkisg | ip a | 11:05 |
| alkisg | Got it; it needed `dmesg -n 8` to actually transfer the messages in 5.04+ | 13:12 |
| alkisg | ...or, echo '4 3 1 7' > /proc/sys/kernel/printk, or loglevel=8 in /proc/cmdline etc; but it appears that this only lasts up until journald is started | 13:39 |
| alkisg | Something resets the log level, so one needs to manually run it again after the system completes its boot process | 13:40 |
| TJ- | alkisg: "grep LogLevel /etc/systemd/system.conf" | 13:44 |
| alkisg | TJ-: thanks; I tried settings that to "debug", I rebooted, but /proc/sys/kernel/printk is again reset to "4 4 1 7" | 13:47 |
| alkisg | What I want is to have "4 3 1 7" or "5 4 1 7" after the system boots, so that a simple "date > /proc/kmsg" reaches the server, | 13:49 |
| alkisg | So, I start with passing "loglevel=5" in the kernel cmdline, and that works up until journal starts, where it's reset to 4. | 13:49 |
| TJ- | alkisg: how about a sysctl setting? | 13:49 |
| alkisg | Do you think journald will respect that? I can try... | 13:50 |
| alkisg | loglevel is supposedly the kernel parameter that is equivalent to kernel.printk... | 13:50 |
| TJ- | alkisg: check out journald.conf, it's man-page shows options MaxLevelStore=, MaxLevelSyslog=, MaxLevelKMsg=, MaxLevelConsole=, MaxLevelWall= | 13:52 |
| alkisg | Those sound interesting, let me try... | 13:52 |
| TJ- | alkisg: see "man journald.conf" | 13:53 |
| alkisg | I tried these in the cmdline, but no joy so far: systemd.journald.max_level_kmsg=5 systemd.journald.max_level_console=5 | 13:59 |
| alkisg | I'll try again after a small nap. Ty, later... :) | 13:59 |
| alkisg | TJ-: I put init=/bin/bash, I run `systemctl mask journald; exec /sbin/init`, and the level is still reset to 4. So it's probably some other component that resets it, not journald... | 14:53 |
| TJ- | alkisg: it does sound like a sysctl thing | 14:55 |
| alkisg | Oh, I see sysctl.d/10-console-message.conf has 4 4 1 7, maybe it's that one, looking... | 14:56 |
| TJ- | I did mention that earlier | 14:56 |
| alkisg | Yup that was it. Sorry I didn't realize you said "maybe it's a setting in sysctl.d", I thought you meant "maybe you can set the one you want in sysctl.d" | 14:59 |
| alkisg | And since I already set it in the cmdline, I didn't look into it more | 14:59 |
| TJ- | well, it took you as long to find that as it's taken me to find a missing managed switch on my network! | 15:00 |
| alkisg | Haha, I had a nap too :D | 15:02 |
| alkisg | It took me all Friday to find a loop in the cabling a school though... a teacher removed a utp cable from a pc, put it to his laptop, and when he was done, he put it back to... the switch, creating a loop | 15:03 |
| alkisg | The school lost networking for a week :D | 15:04 |
| TJ- | arghh! | 15:04 |
| TJ- | well in my case I have a netgear GS748TP, used for labs, so it is powered off most of the time. It should be on the management vlan on 10.254.0.253, but I couldn't find it despite all manner of permutations of looking for it on both the management and default vlans, and trying alternate (inc. default) IP addresses, different ports, and so on | 15:06 |
| TJ- | eventually downloaded the Netgeat SmartWizard windows software, ran it under wine, it found the device instantly and confirmed I had the correct IP address... so at that point I added the management subnet to the default vlan and there it was! so it had forgotten it was supposed to be on the management vlan | 15:07 |
| * alkisg got a headache just reading about it :D | 15:08 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!