Alt TextA screenshot of a file manager preview window for my ~/.cache folder, which takes up 164.3 GiB and has 246,049 files and 15,126 folders. The folder was first created about 1.75 years ago with my system
You don’t have to clean your ~/.cache every now and then. You have to figure out which program eats so much space there, ensure that it is not misconfigured and file a bugreport.
<span style="color:#323232;">% du -sh ~/.cache
</span><span style="color:#323232;">1,6G /home/bizdelnick/.cache
</span>
I don’t remember if I ever cleaned it up. Probably a couple years ago when I moved my old HDD to new PC with freshly installed OS. It does not grow accidentally. Only in some very rare cases. As well as some other dirs under ~ and var. If it is a critical system, set up monitoring of free filesystem space. If not, you will notice if it becomes full (I can’t remember when this happened to me last time, maybe ~15 years ago when some log file started to grow because of endless error messages).
Because some users experienced accidential grows like OP had 160 Gbyte. So general advice for linux users can be stated as: Check your ~/.cache every now and then
Critical systems/servers shall better be monitored as you suggest.
Some users experienced accidential growth of /var/log. Some users experienced accidential growth of /var/cache. Some users experienced accidential growth of /var/lib. Some users experienced accidential growth of ~/.xsession-errors. Shall I continue?
Does every user need to begin his day checking all that places? No, he does not. It is waste of time. Such situations are extremely rare. If you are paranoid, check df to see if you have enough free space, and only if it unpredictably shrinked begin to ivestigate which directory has grown.
I don’t get your point. Why should somebody do this every day?
As the experience from other users in this thread, it seems not extremely rare to have an overgrown ~/.cache/ folder. So checking it from time to time is a good advice. If we all do this for a time, and create bug tickets for software which is not cleaning up. Then this problem will hopefully go away with future software releases.
You: It’s a single user system
Also you: Tmpfs would have to be done for every user
And a /tmp/ symlink would have to be created for every user too, so I don’t get your point
Tmpfs is just as easy as making a symlink, but without the filename conflicts between files in ~/.config/ and /tmp/. You just need to add a line to /etc/fstab
<span style="color:#323232;">find /home/ -mindepth 1 -maxdepth 1 -type d -exec mount none {}/.cache/ -t tmpfs -o size=16G </span><span style="color:#0086b3;">;
</span>
for doesn’t work here because it uses spaces to delimit strings, which could cause issues with filenames that contain spaces
You can also create a systemd user service, which is useful if you don’t have root access. The above mount command requires root, but the following doesn’t and is more robust than symlinking to /tmp/:
Isn’t most of what’s in there just filters downloaded from the internet? Python packages, browser cache, etc? Your system confirms you to redownloading everything all the time, no?
Pacman’s cache isn’t in ~/.cache though, it’s in /var/cache. So whatever is taking up this much space isn’t the package manager.
That being said, I think the arch devs should add a config option to automatically delete old packages without having to run paccache manually and have it default to the last 2 versions of a package or so. It can grow quite big over time.
it doesn’t matter if you use paru, yay or heck makepkg if you are compiling packages with hilariously large sources like for example webbrowser (librewolf, brave, ungoogled-chromium, firedragon take each like ~30 GB) without pruning the build cache afterwards
Something I noticed was that in this case it was mostly binary AUR programs taking up the space.
I think maybe since yay/AUR use cloned git repos, and old versions of binaries get stored in the git diff and then add up because different versions of the binary are basically like keeping multiple copies of it instead of just the changes to the source code.
My update script handles mirrors, updates and cleans the cache automatically. I’d definitely recommend creating one. It’s aliased to sysupdate for me and I also check if it’s a debian or arch based distro so the command works on my servers and desktop
I don’t think I’ve posted it before, but here it is. If you use different utilities you’d have to swap those out. Also excuse the comments, I had GH Copilot generate this script
I’ve heard of tools like that, but this works fine for me. This way I’m not dependent on it being packaged for my distro and having to install it through other means. I’m fine running things manually, this is just for convenience
You can use yay -Sc to clean the cache. It’ll also ask you if you want to clean the pacman cache, which I’m assuming you also haven’t cleaned (check the size of /var/cache/pacman).
Thanks for this! I’ve been meaning to start getting into learning more about systemd and making services, this is super detailed and gives me a pretty good starting point!
Did you happen to see which subdirectory was using up this much space? I don’t think I’ve ever seen .cache go above 10GB, so this may be a bug in a piece of software you use.
Looks like yay is storing every previous binary for AUR bin packages (also excuse the unreadable terminal theme, it doesn’t play very well with a lot of TUI apps unless they support custom theming)
You should run yay -Sc from time to time. This cleans a) your pacman cache (which is normally done by executing pacman -Sc) b) your AUR build cache, which is what’s taking up 160GB. But this one seems rather unusual, I use paru (which also has the command paru -Sc), so I can’t really tell if this is normal with yay.
The command also asks you for every directory if you want to delete it or not, so it’s completely save to run that command.
Something I noticed was that it was mostly the binary packages that were taking up so much space, it may be because of how yay stores the programs (does it use git?), the ones that were compiled from source code usually took up the least amount of space, while the binary programs were the ones taking up tens of gigabytes
Indeed, yay utilizes the AUR, which essentially serves as a Git repository for each package. These repositories typically include a PKGBUILD file and a .SRCINFO file, along with possible additional files like patches, desktop, or service files.
For example, take a look at IntelliJ Ultimate: [aur.archlinux.org/cgit/aur.git/tree/?h=intellij-i…]. It contains the .SRCINFO and PKGBUILD, as well as a .desktop file. These files themselves do not occupy much space.
The PKGBUILD specifies the sources for dependencies. For instance:
The PKGBUILD is essentially a Bash script with predefined functions and variables. You can learn more about it here: [wiki.archlinux.org/title/PKGBUILD].
This script primarily downloads and extracts the tar file. In this specific case, it only relocates the files to their intended installation locations, like moving the desktop file to /usr/share/applications.
With such packages, there’s a possibility of wasting significant space since the tar file is downloaded and possibly retained in the cache.
However, other packages, especially those compiled from source, usually involve Git clones. These clones bring the Git repository into a subdirectory of the already cloned AUR package Git repo. Some might also have source tarballs. These types of packages generally do not consume much space in the cache, as they are often just text files, like C source code or Python scripts. These packages frequently rely on external libraries and packages, which are not included in this package’s cache.
While binary packages often bundle all necessary libraries and other components in their source tarballs.
The AUR cache is mostly beneficial if you’re rebuilding the same version or can reuse components from a previous version. For example, a package might depend on a large, static file that doesn’t change often.
In Paru, I’ve enabled the “CleanAfter” option to prevent my cache from overflowing. Given my relatively fast internet speed, redownloading large files isn’t a major concern for me.
You could have a cronjob run something like find /home/user/.cache -type f -atime +30 -delete, which would find files that haven’t been accessed in the last 30 days and delete them. Make sure your home partition is not mounted with the noatime option though.
Ah I was getting it confused. At one point Steam stored everything in ~/.local/share/steam and symlinked ~/.steam to it. Doesn’t appear to be the case on Ubuntu 22.04, though I used to use Debian and grab the .deb from Valve’s website. My bad! :)
No, .cache is similar to a temporary directory (or at least in theory) where important data isn’t supposed to be stored there, instead only temporary files that might speed things up (e.g. images in a browser or thumbnails in a file manager). In this case it looks like all of my AUR packages had their source files cached, which added up over the ~1.75 years that I’ve been running this distro
13,574 totaling 1.7gb, not too bad. Hey OP how do you get to this view? It looks like we both use nautilus but when I select “properties” on the .cache folder it looks different.
I use thunar (with ePapirus-Dark icons which is probably what makes it look like nautilus), I liked nautilus when I used it but thunar has a bit more functionality that I like
seems like a bug in one of rhe programs you’re using.
modt software automatically manages it’s cache…
are you using build caching tools such as Mozilla sccache? These tend to create 20gb+ cache directories, especially if used with debug builds
Add comment