| mason | The specific example I gave is for Debian. It's actually simpler for Devuan because you don't have to dance around to avoid systemd. | 00:00 |
|---|---|---|
| Xenguy_ | I'll have to read up on it some more; I'm just not sure why I'd ever need to use it : -/ | 00:01 |
| buZz | thats ok, many ppl live their whole lives without needing debootstrap :) in fact, most | 00:01 |
| Xenguy_ | Or perhaps I just have a lack of imagination | 00:01 |
| mason | Xenguy_: My primary use is because ZFS isn't natively supported in the installer. | 00:03 |
| Xenguy_ | Oh, I see | 00:04 |
| mason | So this way, I can just install ZFS systems without having to worry about what the installer wants to support. | 00:04 |
| Xenguy_ | There ya go, that makes sense then | 00:04 |
| rwp | The installer very often wants to partition as the installer wants to partition. But not as I want to partition. The installer partman is pretty rudimentary. It's often much easier to do it manually. | 00:10 |
| rwp | So booting a live boot system. Then setting up the partitioning. Which might be simply cloning the working partitioning from another system. Then using debootstrap to pour the bits on. Then doing some of the basic config. Then rebooting. | 00:11 |
| rwp | Depending upon what I am doing I might prepare one basic system in order to get the template. Then tar it up into a base.tar.gz file. Then later live boot. Partition. Unpack the base.tar.gz and have a working system. But it started from debootstrap. | 00:12 |
| Xenguy_ | Interesting, food for thought. Time to hit the road, thanks. | 00:14 |
| onefang | Ah just as I was getting to the end and was about to stick my oar in, Xenguy_ leaves, but Xenguy is still here... | 00:19 |
| rwp | We all must relocate at times. | 00:21 |
| onefang | I use mmdebstrap, I used to use debootstraap, but I like mmdebstrap better. I have this set of scripts I work on, especially this time around did lots of rewriting and stuff. It mmdebstraps to a qemu image, then I run that in qemu and do the rest of the install. | 00:21 |
| onefang | Then I do lots of testing, then roll it out to my desktops and servers. | 00:21 |
| onefang | I use these scripts to install from scratch rather than upgrade to newer releases. It tweaks things a lot to my liking. | 00:22 |
| rwp | debootstrap is like the lowest common denominator for a fresh install. It's at the very bottom of the layers. But the process of unpacking a deb file and running post install scripts and with the awful decision to fsync() everywhere forcing things to run at disk speeds it's a slow process. But done once then the files are just files on disk. I bundle them into a tar.gz file and pour that on instead. It's many times faster. | 00:23 |
| rwp | My PXE network boot environment has bit rotted (many upgrades) and I need to get it back online again. But on my LANs I used to install by PXE booting and then installing from there. Literally in just a few minutes a system could go from power on to fully provisioned and running. | 00:25 |
| onefang | So right now I have a VM running with Daedalus on my super desktop, using real disk not a qemu image, most things are running on it. Likely next week I'll swap it to be the host release, then start up another VM to test server stuff before I update those. | 00:25 |
| onefang | I'm doing lots of "start with whatever is the latest config file, compare to what I built up over the years, adjust instead of copy this time". | 00:27 |
| rwp | I have this infrastructure of scripts which modify the pristine configuration script into what I want to customize. | 00:29 |
| rwp | I install something and it installs a new pristine configuration file. My script wacks it into what I want. | 00:29 |
| rwp | I upgrade from one version to another version and apply -o DPkg::Options::=--force-confnew which installs the package version of the file. My script wacks it into what I want. | 00:30 |
| rwp | By always installing the new package version then it always comes in with the new whatever the new version needs. | 00:30 |
| onefang | Yeah but there's new things, changes, etc. tmux is notorious, this is the first time I have updated to a newer version where it didn't complain about some old config option that no longer exists. | 00:30 |
| rwp | tmux has been pretty bad about sweeping feature changes. | 00:31 |
| rwp | And apache 2.2 to 2.4 was a huge rewriting change too. | 00:31 |
| rwp | But when those huge changes happen then there isn't any real choice as one can't keep the previous version of the file either. | 00:31 |
| onefang | So THIS time I'm studying manuals and thinking about configs. Even changing what I used to do. THIS is why it's taking so long. Well, that and moving house lots. | 00:31 |
| rwp | Doing a good job of it takes a longer time to do it. | 00:32 |
| onefang | Yep. | 00:32 |
| onefang | But if I do well enough this time, next time will be trivial. B-) | 00:32 |
| rwp | It's never trivial but it's 85% better. And that 85% better is well worth it! We only have to deal with the 15% which we have to deal with. | 00:32 |
| onefang | Yep, next time I start with s/daedalus/excalibur/g then test to see what broke. | 00:34 |
| rwp | This is the way. | 00:36 |
| onefang | Also switching some tools this time around. Who says this 63 year old dog can't learn new tricks? B-) | 00:36 |
| rwp | A good pilot is always learning. | 00:37 |
| onefang | Reminds me, I should install some flight simulators. Though not sure about 3D performance under qemu. | 00:38 |
| onefang | Probably a "leave it until you switch the host to the new system" thing. lol | 00:39 |
| paculino | How easy would it be to make file transfers to a usb drive pause for 1 ms between each file, or at certain intervals of time/storage (throttling speed works too)? It overheats when at full speed. | 04:20 |
| gnarface | probably not that hard, especially if it doesn't have to be exactly 1ms | 04:22 |
| gnarface | you could just write a for loop in bash and add sleep to the loop but it only goes down to a minimum resolution of 1s | 04:22 |
| gnarface | i think there was some other sleep tool with higher precision but i forget what it was called | 04:23 |
| gnarface | someone was just talking up delay though, so maybe that can do it | 04:23 |
| paculino | Would I be able to do the bash loop within the aliases file? | 04:25 |
| gnarface | "aliases file?" | 04:25 |
| gnarface | you mean from your aliases definitions of your bash profile? | 04:26 |
| paculino | ~/.bash_aliases | 04:26 |
| gnarface | ah | 04:26 |
| gnarface | i never broke those out into a separate file | 04:26 |
| gnarface | well, i think it might work actually yea, but not sure | 04:27 |
| paculino | It was the default when I started and I never merged them | 04:27 |
| paculino | There were some in the main file that were commented out though | 04:27 |
| gnarface | but worst case scenario you can just create a shell script in ~/bin and uncomment the part that adds that to your user's path | 04:27 |
| paculino | I know I can define something in its own file and make an alias for it | 04:28 |
| gnarface | bash also has a concept of functions | 04:28 |
| gnarface | there's probably a thousand ways to do this | 04:28 |
| paculino | Preferably, I'd like it to be the default transfer method, but I think I'd have to mess with caja itself to do that, so bash it is. | 04:30 |
| gnarface | well | 04:31 |
| gnarface | you want a really dirty way to solve this? | 04:31 |
| gnarface | try using scp | 04:31 |
| gnarface | no special options; in my experience the cpu overhead and something about it's internal disk syncing enforcement usually seems to dodge this issue in other places i've seen it | 04:32 |
| paculino | I only have an issue with this one specific thumbdrive | 04:32 |
| gnarface | (yes, you're allowed to scp between local disk targets even if you're not running the openssh server) | 04:32 |
| paculino | How easy would it be to mess something up that way? | 04:32 |
| gnarface | no harder than with the regular cp command | 04:33 |
| gnarface | just add "s" in front of it | 04:33 |
| paculino | Okay, that sounds good | 04:33 |
| paculino | With doing it in bash, it'd be through cp anyway | 04:33 |
| paculino | Thank you | 04:34 |
| gnarface | no problem, good luck | 04:34 |
| gnarface | there might be a more finished solution somewhere though, my ideas are often outdated | 04:34 |
| gnarface | i've been relying a lot on bash for stuff that has since been replaced with more specialized programs, i just can't think of any for this right now | 04:35 |
| gnarface | dd for example has explicit control over the amount of blocks to write at once, but doesn't actually deal with filesystems so that'd get... tedious | 04:35 |
| paculino | I tried relying on bash alone for really weird custom formating for time, but quickly realized it was slower and not faster like I expected | 04:36 |
| paculino | (I now use poorly written C) | 04:36 |
| gnarface | yea, sometimes just writing a small C program is the right way to go though, that's the way they did it in the old days | 04:36 |
| paculino | I assumed that bash would be best for something to run in the shell only.... it's nearly three thousand times slower | 04:37 |
| gnarface | heh, yea you don't use bash for stuff that's time critical | 04:37 |
| paculino | (That was only like a second, but still) | 04:37 |
| gnarface | even perl would often be a better choice, surprisingly | 04:37 |
| paculino | I tried bash, R, python, and C. With C, I haven't even bothered anything other than an extremely naive solution yet. | 04:38 |
| gnarface | so the bash loop will work though, the only two problems you'll run into are: 1) too many files 2) files with spaces in the names | 04:39 |
| gnarface | for the "too many files" part you just integrate the xargs program into your command (i'd have to read the man page again for specifics) | 04:39 |
| paculino | Generally I'm syncing iso snapshots of my system or 1-5 ebooks of 100-2000 'pages' each | 04:39 |
| gnarface | for the spaces in the names, you just change the IFS environment variable to only accept newlines as delimiters, rather than the default of space, tab, or newline | 04:40 |
| gnarface | (then you just make sure your list command only shows one file per line) | 04:40 |
| gnarface | then you put all that into the ``'s in: for BLAH in `[file list command here]`; do [move one file command here] && sleep; done | 04:41 |
| gnarface | something like that anyway | 04:41 |
| paculino | But scp works too? | 04:41 |
| gnarface | i always have to refresh myself on the syntax and kinda build it up a layer at a time, doing test runs with echoes and stuff | 04:41 |
| gnarface | oh, the scp idea was instead of the sleep command | 04:41 |
| gnarface | instead of the for loop even | 04:41 |
| gnarface | yea, my theory was scp might slow it down enough on its own to just obviate the problem, but i haven't actually tested that theory since they switched to the sftp backend for it, i realize, so ymmv | 04:42 |
| gnarface | with scp you can just pass -r to recursively copy the whole directory | 04:42 |
| paculino | Is the slowing down by default? I'm not seeing anything on options. | 04:42 |
| gnarface | yea, my theory is that the inherent overhead of scp will just slow it down on its own, no throttling options necessary | 04:43 |
| gnarface | it's just a theory | 04:43 |
| gnarface | if you want to be sure, use the for loop with sleep in it | 04:43 |
| gnarface | but stick around, we have a lot of veteran sysadmins around here, maybe someone else can suggest a cleaner solution than any of these | 04:44 |
| paculino | I'll try scp next time, and if insufficient (or I want to procrastinate something between now and then), I will use the for loop | 04:44 |
| paculino | I suppose the proper solution would be to not use a usb drive that is surrounded by metal and as small as possible.... the bulkier ones with plastic don't overheat | 04:45 |
| gnarface | hmm, yea but it's a rare issue and i have to wonder if there's not something driver related wrong... like it's syncing up at usb2/3 speeds while actually only being a usb1 rated flash device, or something like that (i've seen that before too) | 04:46 |
| gnarface | even with the sleep-loop solution, you might still in theory run into overheating problems if the mix of files contains some very large ones | 04:47 |
| paculino | I've found an old reddit post on it, saying the new bulkier plastic one doesn't overheat | 04:51 |
| paculino | So it's this... I'm plugging it into the USB 3 port and it is getting about what Samsung claims. Someone there said to disable power management to it entirely to fix it, but it's been trouble for the last few years apparently | 04:54 |
| gnarface | hmm, any usb2 ports around you could plug it into instead? | 05:29 |
| gnarface | or usb1 even? | 05:38 |
| gnarface | some old hub or adapter might help | 05:38 |
| gnarface | nothing like adding a hardware bottleneck to purposefully enforce throttling | 05:38 |
| paculino | I do have an old hub, but then I just lose access to the mouse and keyboard since those are beside the overheating usb drive... | 06:12 |
| _al1r4d | https://news.ycombinator.com/item?id=40843778 | 13:49 |
| joerg | gnarface: ((sleep min duration)) https://termbin.com/a1jw | 13:56 |
| joerg | messybox iirc can't do fractions of a second though | 14:01 |
| CueXXIII | joerg: there's more going on than a single sleep, shell scripts are bad for precise timing: https://www.oetec.com/pastebin/plain/b4FUmjcF | 14:12 |
| joerg | I know precise timing isn't possible in bash. I didn't think it was a requirement here? | 14:15 |
| CueXXIII | paculino: i would suggets rsync with --bwlimit to see which bandwith your drive can take without overheating | 14:18 |
| CueXXIII | yeah, i hadn't looked at the original problem yet :) | 14:19 |
| joerg | _al1r4d: what the heck? | 14:22 |
| joerg | https://www.qualys.com/2024/07/01/cve-2024-6387/regresshion.txt | 14:22 |
| CueXXIII | i guess openssh is just one of the most attacked programs | 14:23 |
| joerg | https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2024-6387 | 14:24 |
| gnarface | oh, rsync can bandwidth limit, i knew there was something important i was forgetting here | 14:36 |
| joerg | :-) | 14:42 |
| joerg | also also make sure the USB stick is mount -o noatime,nodiratime | 14:46 |
| systemdlete | a few minutes ago, my chimaera host's desktop totally crashed. Yesterday, I upgraded all of my systems for the new kernel, so I am wondering if that might be the source of the issue, before I go hunting. | 20:10 |
| systemdlete | The host itself did not crash (but I rebooted it anyway, just to be safe). | 20:11 |
| systemdlete | Has anyone else experienced a sudden desktop crash since they upgraded to the newest chimaera kernel? | 20:11 |
| systemdlete | (from the repos, of course) | 20:11 |
| systemdlete | unfortunately, rsyslog was not running and I am not sure why. Since rebooting, I see that it was not automatically started either. I'm looking into that now. | 20:16 |
| systemdlete | Apparently, when I rebooted to the new kernel yesterday, rsyslog did not start. That's as much as I know at this point. | 20:16 |
| systemdlete | I did not make any changes to rsyslog config recently (last update to config was early June) | 20:18 |
| systemdlete | One thing I have found: The start and kill scripts for rsyslog are missing in the rc?.d directories. | 20:57 |
| systemdlete | How would they get removed? I have no reason to do this myself. Could this be the result of an upgrade to, say, rsyslog or sysvinit? (grasping at straws here...) | 20:58 |
| systemdlete | I'll put them back, but it is weird that they are missing. | 20:58 |
| rwp | systemdlete, Welcome to the new Debian where they are actively suppressing init freedom. Probably you can find those in the orphan-sysvinit-scripts package? Which release suite is this? | 21:15 |
| rwp | In Debian a lot of packages have been actively removing init scripts. In Devuan these have been getting collected into orphan-sysvinit-scripts. | 21:16 |
| rwp | In daedalus rsyslog still ships a /etc/init.d/rsyslog script. Therefore I assume you are running Unstable/Testing? | 21:17 |
| fsmithred | Is the init script missing, or just the links are missing? | 21:28 |
| rwp | fsmithred, In rsyslog (8.2110.0-2) * Remove SysV init script -- Michael Biebl <biebl@debian.org> Fri, 05 Nov 2021 23:29:45 +0100 | 21:32 |
| fsmithred | Oh! I just checked my excalibur and there are no links, no init script, and no rsyslog installed. I'm not sure how that happened. | 21:33 |
| rwp | syslog-ng installed? | 21:33 |
| rwp | I apparently slurped the init.d script into my own provisioning scripts infrastructure and have been self-maintaining it myself since then. | 21:34 |
| fsmithred | There was no syslog. I just installed rsyslog and got init script and links. | 21:35 |
| fsmithred | all good. | 21:35 |
| rwp | Apparently it takes so much of my time to maintain my own copy that I have totally forgotten that I did that three years ago. | 21:35 |
| fsmithred | 8.2404.0-2devuan1 | 21:35 |
| rwp | Oh! Of course! I am looking at my Debian Unstable system with 8.2404.0-2. My bad! | 21:36 |
| rwp | systemdlete, Double check your system release. Devuan has forked that package and it should be okay in Devuan. | 21:37 |
| fsmithred | yeah, it's good in daedalus, too | 21:37 |
| rwp | Right. I check my daedalus laptop and it is running 8.2302.0-1devuan1 and okay too. | 21:38 |
| fsmithred | same here | 21:38 |
| systemdlete | This was on a chimaera system. Just the links were missing. update-rc.d rsyslog defaults successfully created the links | 21:39 |
| systemdlete | all of this is from stock repo packages, no customized stuff by me | 21:39 |
| systemdlete | rsyslog was certainly installed and running prior to the upgrade yesterday | 21:40 |
| systemdlete | so something happened fairly recently--that's why I am suspicous of this kernel upgrade. | 21:40 |
| rwp | No /etc/rc?.d/{S,K}*rsyslogd symlinks? Strange! That's not normal. But it is working okay for the rest of us. Something isolated to your system then. | 21:43 |
| rwp | Note that if any symlink exists then the configuration is "installed" and won't be touched further. For example if /etc/rc2.d/S??rsyslog is remove then nothing will recreated it as that removal will be seen as an explicit local admin action. | 21:43 |
| rwp | But if ALL symlinks are removed then dpkg assumes that it has not yet been installed and will install all of the symlinks in their pristine state. | 21:44 |
| systemdlete | I have been running rsyslog on that system for months... years even. | 21:44 |
| rwp | And never rebooted? | 21:44 |
| systemdlete | right... seriously? LOL | 21:45 |
| systemdlete | reboot should not remove files | 21:45 |
| systemdlete | not system files anyway | 21:45 |
| systemdlete | (except maybe state files, etc) | 21:45 |
| rwp | No. But rsyslogd won't start at boot time if the symlink is not there to start it. | 21:45 |
| systemdlete | no kidding! | 21:45 |
| rwp | If it was running then the symlink was there to start it. | 21:45 |
| systemdlete | rsyslog was running prior to the update/upgrade yesterday | 21:46 |
| systemdlete | not sure how, and I am looking at system backups now... | 21:46 |
| fsmithred | what version of rsyslog is there now? | 21:46 |
| fsmithred | I'm booting up my chimaera and I'm sure it hasn't been upgraded lately. | 21:46 |
| systemdlete | fsmithred: 8.2302.0-1~bpo11+1devuan2 | 21:49 |
| systemdlete | it looks like S02rsyslog has been missing for a long time... at least since June 9. | 21:51 |
| systemdlete | That's as far back as I have backups. | 21:51 |
| systemdlete | oh wait... | 21:51 |
| systemdlete | I have more backups | 21:51 |
| systemdlete | it will take some time to access them | 21:51 |
| fsmithred | installing now | 21:54 |
| fsmithred | did you check to see that the postinst script makes the links? | 21:54 |
| rwp | systemdlete, Can you upgrade from 8.2302.0-1~bpo11+1devuan2 to 8.2102.0-2+devuan3 in Chimaera? | 21:54 |
| rwp | Erm... I guess that would downgrade you. Sorry. What's the reason the backport is needed? | 21:54 |
| fsmithred | oops. no more links. | 21:55 |
| rwp | fsmithred, So the backport doesn't include the links? Does it include the /etc/init.d/rsyslogd script? | 21:55 |
| systemdlete | fsmithred, this is rsyslog 8.2302 on refracta chimaera | 21:55 |
| systemdlete | rwp: the links are created by running update-rc.d defaults | 21:56 |
| rwp | Chimaera has 8.2102.0-2+devuan3 in its' repository. | 21:56 |
| rwp | systemdlete, Right. But if the init.d script is present then the postinstall should also include that action. | 21:56 |
| systemdlete | I may have upgraded rsyslog some time back to get better support | 21:56 |
| rwp | In the "Automatically added by dh_installinit" section. | 21:57 |
| fsmithred | no init script in the bpo version | 21:57 |
| rwp | I am very happy to live a boring computer life and be able to use the same features that have been available for decades. :-) | 21:57 |
| fsmithred | I will downgrade to the stable version. BTW, the postinst script does have the line to make the links, but it depends on the existence of executable init script. | 21:58 |
| rwp | That's the problem then. Missing init script means no symlinks generated from the postinst. Even if the init script was left behind as an obsolete conffile . | 21:58 |
| systemdlete | some bugs in earlier versions were fixed | 21:58 |
| rwp | We know the solution now. Time for an updated backport with the init script included. Yay! | 21:59 |
| systemdlete | well, that doesn't explain why rsyslog started on my chimaera system, unless I had run the update-rc.d defaults at some point. And that is likely, given that after upgrading to the bpo version, I probably realized they were missing at the first reboot...? | 22:00 |
| systemdlete | idr now. | 22:01 |
| fsmithred | downgrade is not fixing it | 22:01 |
| fsmithred | even though the script is in the package | 22:02 |
| fsmithred | weird | 22:02 |
| fsmithred | ok, I unpacked the deb and moved the initscript then ran dpkg -i again. | 22:04 |
| rwp | Things get weird with /etc/ conffile handling. | 22:05 |
| rwp | And I blame [name redacted] for arguing successfully some really pedantic odd stuff in the bug tickets getting the current odd behavior now. | 22:05 |
| nsttaffo | is there another way to frugal boot into devuan without netboot? | 22:09 |
| rwp | What is "frugal boot"? Is "netboot" a form of PXE boot? | 22:17 |
| nsttaffo | yeah, from what i heard from it | 22:17 |
| nsttaffo | and i think frugal boot means booting without a usb stick | 22:17 |
| fsmithred | there's something called netboot in the installer tree. | 22:17 |
| fsmithred | I'll find it. | 22:17 |
| fsmithred | Take a look here: https://pkgmaster.devuan.org/devuan/dists/daedalus/main/installer-amd64/current/images/netboot/ | 22:19 |
| rwp | Web search turns up: https://mxlinux.org/wiki/system/frugal-installation/ | 22:20 |
| fsmithred | You might need to dig up some debian instructions unless you already know what to do with those. | 22:20 |
| fsmithred | Looks like they're talking about booting a live-iso from the hard drive. | 22:21 |
| fsmithred | Now that they're gone I see they wanted "without netboot" | 22:27 |
| rwp | I thought it odd that they were asking about it but they did not know exactly what it was they wanted. | 22:31 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!