Use RAM for /tmp instead of my SSD

Help Desk General Help
1

I am not sure if I read or if I heard on a podcast or something about using RAM instead of your SSD for /tmp. I searched around a bit and saw this as a partial answer to my question.

In short, I think what I SHOULD be able to do is add this to /etc/fstab:

tmpfs /tmp tmpfs defaults,nosuid,nodev,noexec,noatime,mode=1777 0 0

I have 40 GiB of RAM and unless I am doing a lot with VMs, I barely use it.

My questions are:

  1. Would adding that to /etc/fstab and nothing else force the usage of RAM for my /tmp folder
  2. Should I add a limit to how much it can grow?
  3. Besides setting a drive size limitation, is there any other way to automatically clear out temporary files in a clean and elegant way?

Thanks!

1 Like
2

According to the Arch Wiki:

It is also used for /tmp by the default systemd setup and does not require an entry in fstab unless a specific configuration is needed.

Source: https://wiki.archlinux.org/title/Tmpfs

So probably you don’t have to do anything if you’re on a system with systemd. You can simply check the status by systemctl status tmp.mount, and double-check the unit file in /usr/lib/systemd/system/tmp.mount.

2 Likes
3

I had a /tmp issue on Fedora yesterday where imagemagick had been dumping massive files into it (it can do that if it doesn’t exit properly). I was getting /tmp “out of space” errors and thanks to your post now I know where that space comes from. Thank you. That made the foundation of the rest of this post:

Unless you want to throw some BASH together tmpwatch is pretty good as a cron entry or a systemd job. Aside from the Debian universe it has general repo availability incl’ OpenSUSE.

# Delete anything that hasn't been accessed in a day
/usr/bin/tmpwatch --atime 1d /tmp

Just make sure atime isn’t turned off for the mount, otherwise use --mtime.

cat /usr/lib/systemd/system/tmp.mount | grep 'Options='
# Fedora output:
# Options=mode=1777,strictatime,nosuid,nodev,size=50%,nr_inodes=400k

From the above tmp.mount output, it looks like mine is capped at 50% of my RAM: size=50%
If you want a specific size you could change that to something like size=20G

1 Like
4 
> systemctl status tmp.mount
â—Ź tmp.mount - Temporary Directory /tmp
     Loaded: loaded (/proc/self/mountinfo; static)
     Active: active (mounted) since Wed 2021-10-27 09:36:05 EDT; 3 days ago
      Where: /tmp
       What: tmpfs
       Docs: https://systemd.io/TEMPORARY_DIRECTORIES
             man:file-hierarchy(7)
             https://www.freedesktop.org/wiki/Software/systemd/APIFileSystems
      Tasks: 0 (limit: 4915)
        CPU: 1ms
     CGroup: /system.slice/tmp.mount

Warning: some journal files were not opened due to insufficient permissions.

So, that doesn’t tell me that it is RAM disk or am I just not understanding what I am reading?

5

Now I am wondering, since the systemd module is running, is there a systemd tmpwatch implementation @Ulfnic ?

6

Excerpt from kernel.org’s section on tmpfs.

Everything in tmpfs is temporary in the sense that no files will be created on your hard drive.

7

It tells you that whatever is described in the tmp.mount file, has been successfully executed.
The lines Where: /tmp and What: tmpfs tell you that at the /tmp directory you have a tmpfs (RAM disc) mounted.

8

Looks like it has a timer for cleanup running by default: CentOS / RHEL 7 : how tmpfiles clHow systemd-tmpfiles cleans up /tmp/ or /var/tmp (replacement of tmpwatch) in CentOS / RHEL 7ean up /tmp/ or /var/tmp (replacement of tmpwatch) – The Geek Diary

cat /usr/lib/tmpfiles.d/tmp.conf
# Output w/o comments:

q /tmp 1777 root root 10d
q /var/tmp 1777 root root 30d

My /tmp is emptied of files that haven’t been accessed in 10 days and /var/tmp for 30 days.

To run the cleanup service manually:

sudo /usr/bin/systemd-tmpfiles --clean

To see it’s timer settings, service settings and status:

cat /usr/lib/systemd/system/systemd-tmpfiles-clean.timer
cat /usr/lib/systemd/system/systemd-tmpfiles-clean.service
systemctl status systemd-tmpfiles-clean
9

Very cool. I learned something here. So, it looks like /tmp as well as some others folders are indeed using tmpfs

tmpfs            20G   80M   20G   1% /dev/shm
tmpfs           7.8G  2.4M  7.8G   1% /run
tmpfs            20G   12M   20G   1% /tmp
tmpfs           3.9G   14M  3.9G   1% /run/user/9000

It also looks like it clears out after 10 days

> cat /usr/lib/tmpfiles.d/fs-tmp.conf
q /tmp 1777 root root 10d

I am not so concerned about the time between clearing out, so long as it does happen. It should also be noted, I do reboot this laptop about weekly, after I sudo zypper dup the thing. My biggest concern was that I was making unnecessary writes to the disk causing premature wear on the NVME.

Thanks everybody. I appreciate the input as I learned a thing or three!

1 Like
10

Same here, thanks Nate

1 Like