| systemdlete | has the output of luksDump changed in the last few years? I'm looking at https://www.maketecheasier.com/change-luks-encryption-passphrase/ and trying the luksDump operation shown about 1/2 way down that page. I don't get the same results. So I wonder if maybe I am unwittingly in a different Linux universe. | 06:37 |
|---|---|---|
| systemdlete | that article was (apparently) published in 2019. | 06:37 |
| AlexLikeRock | whats itsthe PAKAGE to AUTOMOUNT? | 16:07 |
| Stargoose2 | Hi! I am making some tests on a minimal system without sysvnint: I bring everything up "manually" with a custom script I pass to the kernel with the init=... parameter. | 16:13 |
| Stargoose2 | Works great, except... how can I specify what to do at CTRL+ALT+SUP? (three-finger-salute) | 16:13 |
| Stargoose2 | Normally it would be specified on /etc/inittab, but I believe /etc/inittab isn't used in such a system | 16:14 |
| error144 | Hello everyone, hope everyone is fine, I have been trying to use devuan installer-iso to make my own custom distro. | 16:45 |
| error144 | Could anyone tell me where to put the udeb files in order for the 'make' commend to work? | 16:45 |
| error144 | Is there anyone beside rrq that uses installer-iso? | 17:30 |
| error144 | I am shy to ask for help from the same guy 3 times. | 17:30 |
| golinux | FYI . . . rrq CREATES the isos for Devuan so top of the food chain. He' | 17:35 |
| golinux | s still sleeping at this hou. | 17:36 |
| golinux | hour | 17:36 |
| golinux | Come back in 6 hours or so | 17:36 |
| hyrc | "get the rrq" "rrq's sleepin" | 17:57 |
| error144 | golinux I guess it is really hard to do this, I am getting mad, I spent around a week just searching for a way to add an installer for my distro | 19:44 |
| golinux | error144: You've take a deep dive so not surprising you're getting the bends. :) | 19:49 |
| error144 | golinux is it like making an installer for a custom distro that rare? linux has hunderds of distros, how did they make their own iso? | 19:50 |
| golinux | It's not easy to do what you're trying to do so requires patience. | 19:51 |
| golinux | Have you tried refractasnapshot? That's how the live Devuan isos are created. Eve I can manage to do that! | 19:51 |
| error144 | refractasnapshot? I recall rrq said it was more like debian installer then anything else? | 19:52 |
| error144 | yet I found the two way too different | 19:52 |
| error144 | also remember I am trying to make an installer, not a LiveCD | 19:53 |
| golinux | https://www.refracta.org/docs/readme.refractainstaller.txt | 19:53 |
| error144 | !!! | 19:54 |
| golinux | refractainstaller installs exactly what's on the live CD including any changes you might make in the live session. | 19:54 |
| golinux | It is in the Devuan repos/ | 19:55 |
| error144 | I got to try that NOW, thank you golinux | 19:55 |
| error144 | I will be back after trying it. | 19:55 |
| golinux | Good luck! | 19:55 |
| error144 | <3 | 19:55 |
| hyrc | why does devuan ship with kernel setup so that system locks up when OOM, instead of killing offending process? | 20:06 |
| hyrc | maybe this is tradition? | 20:07 |
| hyrc | has been this way in linux since like 18 years, but kernel can be configured to kill the process instead of locking up the whole machine | 20:07 |
| hyrc | maybe changing the defaults would be bad, but on install, end-user should be given the option to configure kernel to not lock-up system when out of memory | 20:08 |
| hyrc | for the desktop user scenario | 20:08 |
| hyrc | thoughts? | 20:09 |
| UsL | ah, you lost the anus. But I agree that OOM should not result in a system lock. | 20:11 |
| hyrc | there might be downsides to my solution that make it bad/wrong for some use cases | 20:13 |
| hyrc | but i much prefer seeing firefox killed than pull power and lose my work | 20:13 |
| hyrc | tell me of your homeworld, UsL :) | 20:15 |
| UsL | are you still banned from -offtopic? | 20:17 |
| UsL | I think that was ridiculous btw. | 20:17 |
| UsL | being banned. Not my home world. | 20:17 |
| hyrc | maybe. joke was from DUNE ("tell me of your homeworld, Usul") | 20:18 |
| UsL | Ah, thought I recognized it. | 20:19 |
| UsL | the old nick is still banned. We can continue in there though. | 20:21 |
| user____ | hyrc: turn off swap and overcommits. | 20:22 |
| user____ | Also oomd is an abomination spawned by systemd & co to live with overcommits. The correct way is to set ulimits per user or per process (pgroups). | 20:23 |
| hyrc | did that yes user____ | 20:25 |
| hyrc | have 3.7GB. setup zram. not sure what % of RAM to use for zram. | 20:26 |
| user____ | If you have ulimits set properly and no swapping (slows down), and reserve 1G or so for kernel, it will not lock up. | 20:26 |
| hyrc | some use 25% some use 199% | 20:26 |
| user____ | zram is not normal ram | 20:26 |
| hyrc | yes it is compressed ram. but what to set it to, idk. seems to work how i have it setup | 20:27 |
| user____ | And yet you complain about crashes. 25% is mentioned. It is bull. Turn off zram, stabilize the system, then turn it back on, one step at a time. | 20:28 |
| hyrc | i fixed them, sir | 20:28 |
| user____ | Who? | 20:28 |
| user____ | 25% ram for zram means something compressed in zram can expand easily to 400% ram and you'll crash for sure. | 20:29 |
| hyrc | i suggested shipping a devuan installer with an install time option presented to user, to not crash the system when memory runs out | 20:29 |
| hyrc | that's how they've set it up for the droid4 with one gb. i used 199% on my 3.7GB machine | 20:30 |
| user____ | Right, that requires black magic and worse. Just do the above. No zram, stabilize the system using cgroups and ulimits, test it is solid no matter how many tabs you open in ff, then gradually enable zram but only for system tmp else ff will use tmp zram and blow up the system again. | 20:30 |
| hyrc | i do not put /tmp on ramdrive | 20:30 |
| user____ | I hope I make myself clear. | 20:30 |
| hyrc | i think you did thanks | 20:30 |
| hyrc | calling it tmpfs was a mistake | 20:31 |
| hyrc | led naive developers to not treat it like a limited-space ramdrive | 20:32 |
| hyrc | (python again) | 20:32 |
| user____ | It is not "limited space", that's the problem. It's dead easy to write a non sparse 60GB file using dd if=/dev/zeror to zram which will be compressed nicely, and completely blow up the system when opened by any app not mapped to zram, using /tmp for temp decompression images, for example. | 20:33 |
| user____ | zram is a way to fake ram which isn't there. You like to live dangerously, do it. But don't complain. | 20:34 |
| hyrc | my lockups didn't occur with my fixes enabled | 20:34 |
| user____ | Okay. | 20:34 |
| hyrc | they occured with devuan as-configured-by-default | 20:34 |
| hyrc | it might be helpful to offer a few options for memory handler on install | 20:34 |
| user____ | yes, but maybe as a patch/post install option? | 20:35 |
| hyrc | you seem to know a lot about this. what do you think of that idea? | 20:35 |
| hyrc | mhm | 20:35 |
| user____ | Arch has nice info on handling low ram https://wiki.archlinux.org/title/Improving_performance | 20:35 |
| user____ | Search overcommit @ | 20:35 |
| hyrc | yeah disabling overcommit helped | 20:35 |
| nemo | hm. I'm confused | 20:35 |
| nemo | tmpfs *isn't* a limited space ram drive | 20:36 |
| user____ | I like uresourced @link, not tried yet | 20:36 |
| nemo | that's what ramfs is for no? | 20:36 |
| nemo | tmpfs was designed to overflow ram if necessary I thought | 20:36 |
| hyrc | tmpfs used to be called ramfs | 20:36 |
| user____ | sigh | 20:36 |
| nemo | hyrc: both are still supported | 20:36 |
| nemo | hyrc: ramfs is ram only. | 20:36 |
| hyrc | then they called it tmpfs and led to tmpfs abuse | 20:36 |
| nemo | hyrc: you can still mount -t ramfs | 20:36 |
| hyrc | maybe i'm wrong about this | 20:36 |
| hyrc | the way things fill up /tmp, i put it on a ssd | 20:37 |
| user____ | tmpfs and zram are somewhat orthogonal. | 20:37 |
| hyrc | and no more angery | 20:37 |
| nemo | if you do dd if=/dev/zero of=/tmp/whatever bs=1G count=100 | 20:37 |
| hyrc | yes | 20:37 |
| nemo | it will work | 20:37 |
| hyrc | we have switched topics | 20:37 |
| user____ | nemo: only with zram | 20:37 |
| nemo | user____: erm. pretty sure it should work without compression. spills over to disc | 20:37 |
| nemo | lemme test locally | 20:37 |
| user____ | nemo: no. Turn off swap first. | 20:37 |
| nemo | user____: ahhh with swap off | 20:38 |
| nemo | yeah. that will fail | 20:38 |
| user____ | https://gitlab.freedesktop.org/benzea/uresourced | 20:38 |
| nemo | user____: ramfs does not use swap | 20:38 |
| user____ | Even with swap on, it will fail. | 20:38 |
| nemo | well. depends how much swap you have | 20:38 |
| nemo | or at least... I thought.. | 20:38 |
| user____ | No normal system has 100G of swap | 20:38 |
| nemo | so... traditionally it was 2X RAM | 20:39 |
| nemo | I guess that's no longer the case | 20:39 |
| nemo | that was back when people used suspend to swap I suppose | 20:39 |
| user____ | Most people do not want to swap at all. The speed penalty is horrible. | 20:39 |
| user____ | Normal swap setup is 1xRAM size normally now. | 20:39 |
| nemo | user____: yeah. one nice thing about ramfs there if you are trying to do something that you absolutely always want to be ram backed | 20:39 |
| nemo | user____: can be helpful for big data files and using shell as data trawler | 20:40 |
| nemo | or... I guess password stuff maybe. dunno | 20:40 |
| user____ | It seems uresourced uses systemd ... | 20:40 |
| nemo | hm. this work laptop has 37.3GiB of swap | 20:42 |
| nemo | what an odd number | 20:42 |
| user____ | How much RAM | 20:43 |
| nemo | thinking someone just grabbed a random rounding off of end of disc when partitioning | 20:43 |
| hyrc | thanks for this informative conversation | 20:43 |
| nemo | user____: 32 gigs, so, ok, about 1X | 20:43 |
| user____ | hyrc: you can script a bash script which implements the changes (and unimplements them) and upload it in the wiki. | 20:43 |
| nemo | definitely would be problematic for hibernating, but I guess no one does that anymore | 20:43 |
| nemo | well. "maybe" | 20:43 |
| nemo | user____: that said, if 1X backing is standard now, the brand new Apple laptop would not fail on count=100 then 😉 - it has 64GiB of RAM so 128GiB RAM+swap if one was going for that ☺ | 20:46 |
| nemo | but yes. was just a randomly typed command. my bad. | 20:47 |
| user____ | Apple does not count for Linux. It's FreeBSD based >;-) | 20:47 |
| user____ | Btw FreeBSD is fairly robust wrt ram abuse by user apps. | 20:47 |
| user____ | It's just that Linux reinvents everything that already exists in *bsd in triplicate, broken by design, and with crap abandonware "docs". | 20:48 |
| nemo | step ①user____ buys me a $4½k laptop step ② I install linux on it. step ③ dd count=100 for some reason ☺ | 20:48 |
| nemo | user____: freebsd doesn't allow overcommit? | 20:49 |
| user____ | Sorry, I retract that. The "optimistic memory allocator" is also implemented in FreeBSD now. Solaris has no overcommits. | 20:49 |
| * user____ headbangs a bit | 20:49 | |
| nemo | well. it probably doesn't have tmpfs anyway | 20:50 |
| user____ | Awe and wonder. I have not touched Solaris since the 1990s, and even then, only as user. | 20:50 |
| user____ | https://www.reddit.com/r/linux/comments/2zbsox/do_i_need_swap_space_if_i_have_16_gb_of_ram/ good reading | 20:50 |
| nemo | I'd rather have it than not have it | 20:51 |
| nemo | esp on a machine used for browsers and java VMs | 20:51 |
| nemo | better to swap thrash than crash | 20:51 |
| user____ | It's your choice. Linux did not invent ramdisks. They exist everywhere but linux renamed it... | 20:51 |
| nemo | well. can still use ramfs if you wish ofc | 20:51 |
| user____ | nemo: swap trashing just delays the swap, | 20:51 |
| nemo | in fact probably on /tmp | 20:51 |
| user____ | *just delays the crash | 20:52 |
| user____ | With the extra bonus of having an unresponsive machine swapping where it is not clear if it will "make it" or not. | 20:52 |
| hyrc | thanks for that link | 20:52 |
| user____ | Idiocy from link: "One significant thing that swap space is used for is tmpfs." -- er, no. Once tmpfs swaps, it is SLOW. | 20:54 |
| nemo | well... why not just use ramfs for /tmp then | 20:54 |
| user____ | tmpfs IS ramfs with another name and other options | 20:54 |
| user____ | such as the ability to swap (and slow down) | 20:55 |
| nemo | the other day I unpacked a year's worth of logs to /tmp to trawl for some data for a user. this was maybe not the best idea but, eh, probably also not that big a deal either | 20:55 |
| nemo | for one thing, the disc itself is virtual on those servers | 20:55 |
| hyrc | it's cartoon level superman-bizarro-world insanity for me to be doing 'desktop-y things' and need gigs of ram. my nextstation hat 24MB RAM. | 20:55 |
| nemo | user____: I'm just saying, why not be explicit and use ramfs if you never want to allow swap | 20:56 |
| user____ | nemo: tmpfs IS ramfs with extras... | 20:56 |
| nemo | hyrc: heh. my daughter is doing school on the living room machine. It is burning 9 gigs of RAM for her friends minecraft worlds. When she fires up minecraft that's another 8 gigs | 20:56 |
| hyrc | and python loves to dump gigs of data in there because "it's temp!" | 20:56 |
| nemo | user____: yes... I'm just saying. if you don't want the extras. just be explicit and mount as ramfs. no more problems right? it'll just fail due to lack of space when it runs out of ram | 20:57 |
| user____ | nemo: and maintain an out of tree config on your own? | 20:57 |
| nemo | user____: I've made ramfs mounts when playing with data files just to avoid swap before | 20:57 |
| nemo | user____: *shrug* it's fstab. fstab is custom anyway usually | 20:57 |
| user____ | https://www.psce.com/en/blog/2012/05/31/mysql-oom-killer-and-everything-related/ more fun | 20:57 |
| nemo | hyrc: it's funny though. up until 2010-ish I had a gentoo laptop with a whole 96MiB of RAM ☺ - it ran an XFCE4 desktop with seamonkey reasonably well | 20:59 |
| nemo | hyrc: P166 - only machine I ever used -Os on, on theory it might help with the RAM situation | 20:59 |
| nemo | hyrc: that's also how much RAM the first generation iphone had | 20:59 |
| nemo | well. I think 128, but some was allocated for graphics. I remember we had a devil of a time getting hedgewars to work on it anyway | 21:00 |
| FilipZ | Hi. | 21:00 |
| FilipZ | Some time ago, I reported here my problem with the malfunctioning wireless card | 21:00 |
| hyrc | nemo: i didn't realize 1st gen iphone was 128MB. I thought 256 was low on N900. | 21:01 |
| nemo | hyrc: https://en.wikipedia.org/wiki/IPhone_(1st_generation) | 21:02 |
| hyrc | hi FilipZ | 21:02 |
| user____ | hyrc: what was your `cat /proc/sys/vm/panic_on_oom` value before changes, and now? | 21:03 |
| hyrc | cat: /proc/sys/vm/panic_on_oom: Cannot allocate memory | 21:04 |
| FilipZ | It was Intel Wireless 3650. I wrote about it under the name of Guest11 and Guest10. | 21:04 |
| hyrc | don't recall changing that | 21:04 |
| hyrc | maybe i did have to check notes | 21:05 |
| user____ | hyrc: try with sudo | 21:05 |
| user____ | cat /proc/sys/vm/panic_on_oom -> 0 or 1 | 21:05 |
| user____ | Is 0 here, default current Beowulf kernel | 21:05 |
| FilipZ | If somebody out here that helped me could remember and tell what could be done next, it would be nice. | 21:05 |
| hyrc | should work as root (su'd) | 21:05 |
| user____ | yes | 21:05 |
| FilipZ | I pasted here a link to the dmesg before. | 21:05 |
| user____ | something strange is going on there hyrc | 21:05 |
| FilipZ | I remember that I was asked about something like running the live OS, in some special mode, in order to do the dmesg once again | 21:06 |
| user____ | https://sysctl-explorer.net/vm/panic_on_oom/ | 21:07 |
| hyrc | FilipZ: i'm sorry i don't have a log of your conversation | 21:07 |
| hyrc | ty user____ | 21:07 |
| hyrc | i set vm.oom-kill = 0 | 21:08 |
| hyrc | vm.overcommit_memory = 2 | 21:08 |
| user____ | Was oom_kill not 0? | 21:08 |
| hyrc | dunno | 21:08 |
| golinux | FilipZ: This channel is logged at http://reisenweber.net/irclogs/libera/_devuan/ | 21:09 |
| user____ | cat /proc/sys/vm/oom_kill_allocating_task is 0 here, should set to 1 to zap processes which try to bust the available remaining ram | 21:10 |
| user____ | That alone sould prevent most kernel panics due to low ram | 21:10 |
| hyrc | copied to notes, ty | 21:10 |
| user____ | https://www.kernel.org/doc/Documentation/sysctl/vm.txt see here original docs | 21:10 |
| hyrc | after closing firefox i get 0 user____ | 21:11 |
| hyrc | so i guess 'cat' was being killed :) | 21:12 |
| user____ | yeah | 21:13 |
| user____ | you are living on the edge | 21:13 |
| user____ | what's your `cat /proc/sys/vm/overcommit_memory` ? | 21:13 |
| hyrc | 2 | 21:13 |
| critr | it's the difference between rebooting immediately and waiting until you can't deal with slow swap and rebooting anyway. | 21:15 |
| user____ | Ok, so you turned it off. Default is 0. Ok as you set it. | 21:15 |
| user____ | critr: you can't reboot when the system is frozen in a swap frenezy. | 21:15 |
| user____ | admin_reserve_kbytes is also interesting hyrc, you want it enlarged | 21:16 |
| user____ | 8MB is not enough for a gui login and process kill. | 21:16 |
| user____ | It is enough for a text login, if you manage to switch to text console while the system is swap trashing. | 21:17 |
| user____ | I'd set oom_kill_allocating_task and be done | 21:22 |
| user____ | hyrc: so set /proc/sys/vm/overcommit_ratio to 0 and turn off swap and it should "never" crash again | 21:32 |
| user____ | The kernel. | 21:32 |
| hyrc | is the overcommit ratio still relevant with overcommit_memory = 2 ? | 21:33 |
| hyrc | it's 50 currently | 21:33 |
| user____ | Honestly, I go with defaults whenever I can. Only if things go South do I bother. | 21:33 |
| hyrc | :) yes | 21:33 |
| user____ | https://docs.kernel.org/admin-guide/sysctl/vm.html @ overcommit_ratio | 21:34 |
| hyrc | thanks | 21:34 |
| user____ | hyrc: see @ CommitLimit for formula used https://www.kernel.org/doc/html/latest/filesystems/proc.html | 21:34 |
| hyrc | saved | 21:36 |
| FilipZ | I am back again. http://reisenweber.net/irclogs/libera/_devuan/_devuan.2021-09-17.log.html Here are the last mentions of this problem. I am Guest11 here. gnarface: Can I somehow run this safe mode boot entry, without a live OS? | 23:23 |
| gnarface | FilipZ: yea, afaik "safe mode" should be just some kernel command-line options. if you hit "e" to edit the grub boot prompt you can read them and copy them manually to your other grub install | 23:24 |
| gnarface | safe mode options should have been generated automatically by the stock kernel packages but if you put a custom kernel in there maybe not | 23:25 |
| systemdlete | I'm looking at https://www.maketecheasier.com/change-luks-encryption-passphrase/ and trying the luksDump operation shown about 1/2 way down that page. I don't get the same results. Article is from 2019. Anyone know what is going on? | 23:43 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!