IRC channel logs

2018-08-10.log

back to list of logs

<lfam>What filesystem is it?
<emacsomancer>ext4
<emacsomancer>I think I'm going to reinstall on a different drive, but I was curious if I should expect greater inode usage with guixsd, and so whether I should provision extra inodes
<emacsomancer>(I think I only have the choice between ext4 and btrfs on guixsd, and ext4 seemed the safer choice)
<lfam>I've seen this issue before. You should create it with more inodes than whatever it did by default. I'm curious, how large is the filesystem and how many inodes does it have?
<emacsomancer>It's a 50Gb partition with 3.2million inodes, but /gnu is trying to use 7million something (I didn't install an extraordinary number of packages or anything)
<lfam>What is the inode size?
<lfam>That does seem like a lot, by the way
<emacsomancer>lfam: I wonder if something got borked with my guixsd install - I'm not sure if the lack of inodes is the cause or a symptom.
<emacsomancer>inode size is 256
<lfam>I misread your problem originally. I thought you simply ran out of inodes, which can happen with the ext4 defaults sometimes. But you are sying the filesystem is using more inodes than exist? Can you share the error message?
<lfam>I think that 3 million inodes is way more than enough to get started
<emacsomancer>when I do `guix gc` I get the error: `guix gc: error: build failed: setting synchronous mode: unable to open database file`
<emacsomancer>and guix gc --verify=repair,contents doesn't seem to help
<lfam>That sounds like something is wrong the the store database, which is located in /var/guix/db
<emacsomancer>so you don't think that's connected with the inode issue? (having no inodes seems to mess up lots of things, I'm unable to open another emacs --daemon, to run `screen` etc.)
<lfam>I'm confused about the inode thing. Does `tune2fs` report that you are out of inodes?
<lfam>`tune2fs -l /dev/$filesystem`
<emacsomancer>it claims I have 952 free inodes
<emacsomancer>but if I try to run `screen` it says "no space left on device", but there are 12G free
<emacsomancer>so I assume that's "not enough inodes"
<amz31>yes inode issue looks like at least
<amz31>well, another reason might be broken hd
<emacsomancer>could be. but it's a pretty new ssd.
<amz31>it's new or used?
<amz31>ho
<amz31>then forget
<amz31>last time I had this error, someone told me to rune smarttools
<amz31>smartmontools 6.6 out gnu/packages/admin.scm:1481:2
<emacsomancer>I'm planning to reinstall on a larger drive anyway, I think. I'm just worried about whether I just provise extra inodes or (shudder) try btrfs instead
<emacsomancer>amz31: no inodes means I can't install new packages either
<CompanionCube>maybe XFS?
<amz31>emacsomancer: last time some one was complaining about btrfs or something that looks like that
<emacsomancer>amz31: I don't understand
<lfam>3 million inodes should be enough
<emacsomancer>CompanionCube: only ext4 and btrfs are supported by guixsd, I think
<lfam>I wonder what is using all the disk space? A new install used 38 GB?
<amz31>how did you a create that partition?
<emacsomancer>lfam - I'm not sure. this was after using it for a few days, and I installed some (but not an extraordinarily high number of) packages
<lfam>That really seems like too much
<emacsomancer>amz31: using fdisk or whatever the recommended tool in the guix guide was
<amz31>hm
<emacsomancer>(cfdisk)
<amz31>them use some mk.ext4 command or something?
<emacsomancer>yes, and then mkfs.ext4 ....
<amz31>well, my point is that call might be mis configured
<emacsomancer>so nothing weird, in fact just what the guixgd recommended
<emacsomancer>guixsd
<lfam>I have a GuixSD workstation with a 100 GB ext4 partition that contains most of the system. About 60 GB is historical /gnu/store, over a couple years. It's only using 1,289,223 256-byte inodes
<emacsomancer>lfam: so it doesn't sound like I *should* need to over-provision inodes.
<pkill9>wow lfam
<lfam>No, I've only seen that required when created small ext4 filesystems. The defaults in that case can be exceeded in practice
<lfam>Can you share your config.scm?
<pkill9>so you haven't removed old generations and run the garbage collector for the past couple years?
<amz31>maybe misconfigured ext4 during mkfs.ext4
<lfam>pkill9: Not in a while
<lfam>Sometimes to test things
<pkill9>guix is so nice for offloading builds
<emacsomancer>lfam: https://pastebin.com/4ePznwgS
***ata2001_ is now known as ata2001
<ata2001>hello
<ata2001>I installed guix, but fontconfig's weight property doesn't seem to work for some reason
<ata2001>st -f "DejaVuSansMono:pixelsize=14:antialias=true:hintstyle=hintslight:weight=demibold"
<ata2001>st: can't open font DejaVuSansMono:pixelsize=14:antialias=true:hintstyle=hintslight:weight=demibold
<ata2001>it works fine without the weight property
<ata2001>the same command works on my debian and gentoo systems
<pkill9>where do you have the fonts installed ata2001?
<ata2001>at ~/.guix-profile/share/fonts/truetype/
<ata2001>I ran guix package -i font-dejavu
<pkill9>the solution may be to add that path to your fontconfig user configuration, but i cna't remember
<ata2001>it loads the font if I leave out the weight property btw
<pkill9>ok, idk
<pkill9>could try running `fc-cache -f`, but you probably already have
<ata2001>didn't help
<pkill9>(actualy the path thing was for if you put fonts in a place other than your system profile or your user profile)
<ng0>we might be missing font properties
<ng0>other systems install those, our font installations are quiet young
<ecbrown>i'd like to try out guixsd on my macbook pro (macbook11,5) but when i boot, i get some "rich" textual interface that never seems to yield the login
<ecbrown>is there a way to force an "ugly" console, eg. with grub config?
<ecbrown>maybe this is vesa
***nlsun is now known as Guest15279
<database>how to remove gnome desktop
<mange>You should be able to just remove gnome-desktop-service from your config.scm and reconfigure your system.
<database>mange: okay