KatolaZ | no fsmithred | 00:00 |
---|---|---|
KatolaZ | live iso is too big | 00:00 |
KatolaZ | netinst is a few hundred MBs | 00:01 |
fsmithred | too big for what? | 00:01 |
specing | Would anyone be kind enough and add compress=lzo, compress=zlib, autodefrag and nossd to btrfs filesystem mount options in the installer? | 00:01 |
KatolaZ | if we add live iso it will grow to more than 1GB... | 00:01 |
fsmithred | oh, I'm talking about putting that image on usb | 00:01 |
KatolaZ | specing: why? | 00:01 |
KatolaZ | (I mean, why should that be in the installer?) | 00:01 |
specing | KatolaZ: because the compress flags are used for new files only -> making install use less space if they are applied right away | 00:02 |
specing | also make it use mkfs.btrfs --mixed for devices under 16GB | 00:02 |
specing | KatolaZ: also related, is there an easy way to pause the installer right after setting up the partitions but before install base system runs? | 00:03 |
KatolaZ | specing: an installer is a generic tool is not a tool to install just *your* system | 00:03 |
specing | KatolaZ: in general those flags are good to have on a btrfs filesystem | 00:03 |
KatolaZ | specing: just use the expert install option | 00:03 |
specing | not just on mine | 00:03 |
specing | ok | 00:03 |
specing | Why isn't the partitioner more like this? https://linx.li/logicalpartitioner.png | 00:30 |
MinceR | afaik you can use multiple partition-like objects (even on multiple disks) as a single area for LVM on which to make logical volumes | 00:31 |
specing | that is true, the same also holds for btrfs | 00:32 |
specing | it is also confusing that they don't follow in logical order of setup | 00:33 |
specing | SCSI - LUKS - LVM | 00:33 |
specing | instead it is LUKS - LVM - SCSI | 00:33 |
specing | Hahaha now I made it without that in-target already running and OOM kills apt-get | 01:16 |
specing | :( | 01:16 |
specing | running error* | 01:16 |
specing | I don't get it. The VM has 380MB RAM. apt can't possibly be using so much RAM, can it? | 01:22 |
specing | The only explanation would be if for some reason it started installing to tmpfs | 01:23 |
specing | but df -h does not indicate that | 01:23 |
specing | why is it mounted at /dev/.static/dev? | 01:25 |
gnarface | specing: without swap? | 01:28 |
specing | okay, there were 350 MB that the partitioner left in /etc and /media that I temporarily copied to /tmp in order to reformat btrfs | 01:29 |
specing | and forgot to wipe it after copying back | 01:29 |
specing | gnarface: yes | 01:29 |
specing | Well, that is great | 01:36 |
specing | as soon as I have to start killing processes to restart the install at a previous step, something gets noted somewhere and then I get "in-target already running" errors | 01:37 |
specing | bah oom again | 02:56 |
gnarface | there is actually some apt config value you can use to tell it to be sparing on ram | 03:00 |
gnarface | but it's easier to just have swap | 03:00 |
gnarface | you're gonna keep running into this issue otherwise | 03:01 |
specing | I gave it 1.5 times the ram that my Gentoo-running router has, come on :P | 03:03 |
gnarface | well, gentoo works differently | 03:03 |
specing | portage is insanely cpu and ram hungry | 03:04 |
specing | oh wait, does devuan install cd have zram? | 03:05 |
gnarface | i don't know | 03:05 |
gnarface | but i DO know the stuff that i've told you i know | 03:10 |
gnarface | at this point you're just making it harder on yourself | 03:11 |
gnarface | apt started using lots of ram by default a few major debian revisions ago (sometime around when they ditched the "3 floppy disks" version of the installer and the 8MB RAM requirement) | 03:15 |
gnarface | you'll find that the amount of ram it takes is now some factor of the amount of packages pending updates that it's trying to download concurrently | 03:16 |
gnarface | but like i said, in some man page there's a config flag you can change to tell it to be sparing. it'll be slower and try not to unpack everything at once. | 03:17 |
gnarface | but you could probably also just fix this by having like... any swap at all | 03:17 |
gnarface | like 10MB of swap might be enough | 03:17 |
gnarface | your problem is you think of swap as just a linear extension of physical ram, operating at a slower throughput and higher latency, which is a dramatic and harmful oversimplification that's leading you to do insane things like try to omit using swap at all | 03:18 |
gnarface | and not that there aren't valid use cases for a swapless system | 03:19 |
gnarface | it's just that you're clearly not in one of them | 03:19 |
gnarface | and so you're metaphorically trying to bust down a brick wall with your forehead | 03:19 |
gnarface | hell, it might even work to just try to update the packages one at a time. if none of them have a lot of dependencies that could avoid hitting your ram ceiling too, you know. | 03:21 |
gnarface | (just fyi if your complaint about swap is that the system will use some even if ram isn't 100% maxxed out, the correct way to fix that is to set /proc/sys/vm/swappiness to 0, not to just remove swap entirely) | 03:51 |
Leander | do you have any reference about how the swap is more than just a memory extension? this is a comment that comes back again and again in discussions about swap usefulness, but I've never found any detailed technical explanation about why | 03:59 |
gnarface | Leander: i don't have a reference. it was roughly explained to me once in #debian by someone who was just about fed up with my shit too. i'm sure the docs are out there i just don't have them at my fingertips. but the important takeaway is just that omitting swap doesn't really make stuff faster. it superficially makes it seem that way by just preventing anything that needs more space from working at all. | 04:01 |
gnarface | the trick of setting swappiness to 0 will have the same effect on programs that would have worked either way while still allowing stuff that needs more memory to work too | 04:03 |
gnarface | hmm. there should be information on how swap is actually used somewhere in orbit of kernel documentation | 04:04 |
gnarface | think of it in layers | 04:07 |
gnarface | another massive oversimplification probably riddled with inaccuracies but if you think of a cpu with layer-1, layer-2, and sometimes a layer-3 cache.... well just think of system ram as layer 4 and the swap as layer 5 then | 04:11 |
gnarface | removing the swap to make your system go faster is like making your car go faster by removing the brakes | 04:15 |
gnarface | it will go a lot faster | 04:15 |
gnarface | and then crash | 04:15 |
gnarface | the default swappiness value of "60" (out of 100 iirc) is a big swap-happy for the taste of modern users used to having enough ram to fit the entire OS | 04:17 |
gnarface | a bit* | 04:17 |
gnarface | it's really a matter of taste though | 04:18 |
Leander | yes, I can understand the argument that you need swap as a safety net, to give you time to save and/or close things, before OOM rips your system | 04:18 |
Leander | I found interesting answers (that I'd like to investigate further) on that thread: https://unix.stackexchange.com/a/190497 | 04:19 |
gnarface | it will also help with stuff that people benchmarking this stuff often fail to benchmark... like the second time firefox starts up after boot, instead of just the first | 04:20 |
Leander | also https://unix.stackexchange.com/a/190572 | 04:20 |
gnarface | hmmm, yea they look like they know what they're talking about there | 04:21 |
gnarface | now, how much swap you really need is highly debated and i don't even know what the right answer is there | 04:24 |
gnarface | i was still setting mine to the same amount as physical ram up to 8GB, but outside of massive memory leak conditions i've never ever seen it use more than a few megabytes of it | 04:25 |
gnarface | it really depends on what you're actually doing though i suppose | 04:25 |
gnarface | obviously there's certain server tasks that can't work fast enough if they're swapping out (like apache/mysql) and in those cases, use of swap is usually a warning sign, but in none of those cases will just omitting the swap entirely make your life easier | 04:30 |
gnarface | (or even address the actual problem) | 04:31 |
gnarface | (which is probably some rogue php script) | 04:31 |
system32 | Hi. how can i change the timezone ? i have no GUI or DE . also , im using SSH | 13:20 |
gnarface | install the tzdata package (apt-get update && apt-get install tzdata) or dpkg-reconfigure tzdata if it is already installed | 13:22 |
system32 | and after that ? | 13:23 |
debdog | don'T worry, be happy *sing* | 13:24 |
gnarface | system32: after that, follow the onscreen prompts | 13:26 |
system32 | it says tzdata is installed but dpkg-reconfigure tzdata says tzdata is not there | 13:26 |
gnarface | weird | 13:26 |
gnarface | are you fully updated? | 13:26 |
system32 | let me rerun the update command again | 13:26 |
system32 | ah there we go | 13:27 |
system32 | it opened a gui | 13:27 |
system32 | Current default time zone: 'Asia/Tehran' | 13:28 |
system32 | Local time is now: Wed Feb 20 21:22:21 +0330 2019. | 13:28 |
system32 | Universal Time is now: Wed Feb 20 17:52:21 UTC 2019. | 13:28 |
system32 | but thats still incorrect | 13:28 |
system32 | local time is 3:55 pm | 13:28 |
gnarface | you might want ntp too | 13:29 |
system32 | ntp ? | 13:29 |
gnarface | network time sync | 13:29 |
system32 | can i just manually set the time ? | 13:29 |
gnarface | yes | 13:30 |
system32 | well, do you know the command for that ? | 13:30 |
debdog | date -s ..... | 13:31 |
system32 | it worked. thanks | 13:32 |
specing | So | 13:34 |
specing | apparently devuan installer still pulls some stuff from deb.debian.org even though I had not selected that mirror | 13:35 |
gnarface | it should pull most of it from debian.org afaik | 13:38 |
specing | maybe I am doing my proxy mirroring wrong | 13:40 |
specing | maybe I should just make my local DNS server pretend to be authoritative for deb.debian.org and deb.devuan.org | 13:40 |
specing | that way apt would still work after taking machines outside my network | 13:40 |
specing | gnarface: how many domains does devuan pull from? Which ones? | 13:41 |
specing | just deb.debian.org and deb.devuan.org? | 13:41 |
gnarface | well there's a bunch of mirrors in the round-robin | 13:42 |
gnarface | but unchanged debian packages are pulled by redirect | 13:42 |
specing | gnarface: in the public round robin sure, but if I overrode it, then there would be just one physical server in the list | 13:43 |
specing | gnarface: as far as I understood, the redirect happens mirror-side, with alsomething | 13:43 |
gnarface | i'm not sure about that | 13:44 |
gnarface | it's called amprolla3 | 13:45 |
specing | yeah, that | 13:46 |
KatolaZ | specing: you need your proxy to manage the rewrites | 14:03 |
KatolaZ | amprolla is not doing anything online | 14:03 |
KatolaZ | when you contact a devuan mirror, all the redirects are managed by the http server | 14:03 |
KatolaZ | amprolla merges off-line | 14:03 |
specing | KatolaZ: ok, but I'm just forwarding all requests | 14:03 |
* KatolaZ shrugs | 14:04 | |
specing | it is a devuan mirror doing the redirects | 14:04 |
KatolaZ | specing: if other users can use devuan mirrors directly, and you can't use them properly with a proxy, I guess the problem might be with your proxy | 14:04 |
specing | KatolaZ: it works now | 14:04 |
KatolaZ | great | 14:04 |
specing | I was missing a https rewrite to http | 14:04 |
KatolaZ | see | 14:05 |
specing | since my local mirror is only http | 14:05 |
KatolaZ | ok | 14:05 |
specing | and the remote mirror was upgrading to https | 14:05 |
KatolaZ | glad to hear it works | 14:05 |
KatolaZ | specing: that depends on the specific mirror you are contacting I guess | 14:05 |
KatolaZ | if you use deb.devuan.org, you only get HTTP | 14:05 |
KatolaZ | bbl | 14:06 |
specing | Looks like I will not be able to create a DNS-level redirect for deb.debian.org, as ISC BIND does not let me add apex level CNAMEs | 19:30 |
specing | and I don't want to override debian.org | 19:30 |
KatolaZ | specing: if it's not for too many machines, just use /etc/hosts | 19:32 |
KatolaZ | I guess you cannot define CNAMEs for zones you don't manage | 19:33 |
specing | I can pretend to manage them | 19:33 |
DonkeyHotei | a zone cannot be a cname | 19:34 |
KatolaZ | namely | 19:34 |
specing | yep DonkeyHotei | 19:34 |
KatolaZ | specing: you should pretend to manage debian.org and create a CNAME for "deb" | 19:35 |
specing | KatolaZ: yes, but then I have to add all names in the debian.org space myself | 19:35 |
specing | and addresses | 19:35 |
KatolaZ | uh? | 19:35 |
KatolaZ | I don't follow | 19:35 |
KatolaZ | you want deb.debian.org to point to your proxy? | 19:36 |
specing | KatolaZ: if I pretend to manage debian.org, then I have to add A record for debian.org and who knows how many other records | 19:36 |
specing | I don't think there is an easy way to say "if name is not found, query that other server" | 19:36 |
DonkeyHotei | there is | 19:37 |
DonkeyHotei | just use /etc/hosts | 19:37 |
KatolaZ | yeah | 19:37 |
KatolaZ | if it's not for a large number of machines, using /etc/hosts is the best way | 19:38 |
specing | DonkeyHotei: then I have to maintain early /etc/hosts on all machines | 19:40 |
DonkeyHotei | early? | 19:41 |
DonkeyHotei | you're not making sense | 19:41 |
specing | early = to have installation use the new addresses | 19:43 |
specing | better to have an override at the local DNS server | 19:43 |
DonkeyHotei | then override it at the dns server using /etc/hosts | 19:44 |
DonkeyHotei | what's so hard about that? | 19:44 |
specing | are you serious? | 19:46 |
specing | hmm could this actually work, does BIND even query /etc/hosts? | 19:46 |
specing | gotta try | 19:46 |
DonkeyHotei | dnsmasq does, unbound does, why not bind? | 19:47 |
specing | it does not work | 19:48 |
specing | wait, why am I using BIND | 19:51 |
specing | wasn't there some proven correct DNS server written in Ada? | 19:51 |
specing | How much of stuff gets downloaded from deb.debian.org, anyway? | 20:06 |
fsmithred | specing, any packages that we didn't change to remove systemd, so around 99% of the repo | 20:07 |
specing | so adding a proxy mirror for devuan did exactly nothing? | 20:08 |
fsmithred | sorry, I don't know the answer to that | 20:08 |
fsmithred | the 1% of packages from devuan are pretty important | 20:09 |
specing | my devuan mirror cache is 500MB right now | 20:09 |
specing | that includes ascii and testing | 20:09 |
fsmithred | that's just the packages you installed? | 20:09 |
specing | disregard this, I am illiterate | 20:10 |
specing | rewrite /merged/pool/DEVUAN/(.*) http://auto.mirror.devuan.org/devuan/pool/$1; rewrite /merged/pool/DEBIAN-SECURITY/(.*) http://deb.debian.org/debian-security/pool/$1; rewrite /merged/pool/DEBIAN/(.*) http://deb.debian.org/debian/pool/$1; | 20:10 |
specing | ^in amprolla config | 20:10 |
specing | this is actually even better | 20:10 |
specing | as I can rewrite deb.debian.org returned by mirror server | 20:11 |
specing | to my local server | 20:11 |
specing | Will openrc as init work well in ASCII? | 20:16 |
fsmithred | from what little I've seen, it works fine | 20:17 |
fsmithred | all I've done is tried choosing it during installation and also making live isos with openrc. Works in both cases. | 20:18 |
fsmithred | works=boots | 20:18 |
xinomilo | is there a beowulf image for arm (raspi3)? | 20:30 |
golinux | xinomilo: Ask on #devuan-arm | 20:38 |
xinomilo | tried to, but can't go into that channel | 20:38 |
xinomilo | "can't join channel" | 20:38 |
golinux | Then query #parazyd | 20:39 |
golinux | and ask to be allowed in or just ask the question | 20:39 |
golinux | (forget the #) | 20:40 |
specing | Alright, I've added some more stuff to my devuan proxy | 20:41 |
KatolaZ | xinomilo: there is no beowulf image | 20:58 |
KatolaZ | except for the unstable installer for i386, amd64, and ppc64el | 20:58 |
Centurion_Dan | KatolaZ: we should try building debian-installer for arm too. | 20:59 |
KatolaZ | I have tried Centurion_Dan but there are issues with u-boot | 21:00 |
Centurion_Dan | ok... we'll have to fix them then. | 21:01 |
specing | GRUB install failed https://linx.li/s/devuangruberror1.png | 21:03 |
fsmithred | specing, third to last line tells you what to do | 21:04 |
fsmithred | edit /etc/default/grub | 21:04 |
specing | I did that, I am just telling you that you need to fix the installer so this does not happen again | 21:05 |
fsmithred | I thought that worked in ascii. Or are you installing beowulf? | 21:10 |
specing | no, I am installing ascii right now | 21:12 |
specing | for the 10th time | 21:12 |
specing | Looks like 10th time was the charm, I am now booted into a working devuan install | 21:22 |
specing | surprised I did not have to modify fstab for it | 21:22 |
fsmithred | specing, how big is the install? | 21:31 |
specing | fsmithred: just with console utilities, ssh and vim: 565MB | 21:36 |
specing | (btrfs lzo compression - for vim it was turned off though) | 21:37 |
specing | had to edit fstab and I can't stand nano | 21:38 |
specing | also added tmpfs on /tmp | 21:39 |
specing | no idea why this is not default | 21:39 |
specing | I have updated my devuan on btrfs post | 21:49 |
furrywolf | url? | 21:52 |
specing | furrywolf: probably this https://dev1galaxy.org/viewtopic.php?pid=14462#p14462 | 22:27 |
specing | from #d1g-users: | 22:59 |
specing | There is no distro installer that could produce the exact setup I want (btrfs on lvm on luks, with not partition tables underneath). Fedora's anaconda requires 3 GB RAM minimum and cannot even do /boot on btrfs. I don't recall anymore what SUSE's problem was. Devuan's installer apparently insists on having regular partitions underneath LUKS and does not support setting btrfs-specific mount options. So my | 23:00 |
specing | distro installation plan goes as follows: | 23:00 |
specing | - install normally in VM to a disk in tmpfs | 23:00 |
specing | - copy filesystem contents to real disk and fix grub configs | 23:01 |
MinceR | this method might work too >> https://www.debian.org/releases/oldstable/amd64/apds03.html.en | 23:01 |
specing | # mke2fs -j /dev/sda6 | 23:05 |
* specing laughs all the way to /lost+found | 23:05 | |
specing | jokes aside.. | 23:05 |
specing | Looks like a good tutorial, I will save it for some other time | 23:06 |
specing | since I now have a working install obtained by regular methods | 23:07 |
specing | There are two more problems, one on shutdown (with OpenRC): | 23:08 |
specing | https://linx.li/kn2ru2ws.png | 23:09 |
specing | Second when unlocking LUKS in initramfs, the screen is not properly cleared before the password prompt is displayed: https://linx.li/devuanlukstext.png | 23:11 |
specing | or perhaps some log is being printed at the prompt, not sure | 23:11 |
specing | proxy mirroring "howto" updated as well: https://dev1galaxy.org/viewtopic.php?pid=14502#p14502 | 23:36 |
fsmithred | specing, I don't think I've ever seen the screen cleared when asking for the luks passphrase | 23:39 |
Generated by irclog2html.py 2.17.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!