Micha Silver

#gisChat #RStats
On the Debian wiki(https://www.debian.org/releases/trixie/release-notes/issues.html#the-temporary-files-directory-tmp-is-now-stored-in-a-tmpfs) there is a suggestion to revert to the "old" behavior of an actual "/tmp" directory, on the root filesystem (which would normally have much more free space than 8 GB):

"You can return to /tmp being a regular directory by running systemctl mask tmp.mount as root and rebooting."
3/

If anyone has some similar experience, or can shed some light how to solve this situation permanently, I'd be open to suggestions.
7/

But what about the spatial analysis? I still have only 8 GB of temp space. Here's my workaround: R sets up a temporary directory at startup either in the system temp location(i.e. /tmp) or based on the TMPDIR environment variable. So I exported an env variable TMPDIR to a temporary directory in my home dir, then started R. Before running my script I checked 'tempdir()' and sure enough it pointed to that temp directory. Fired off the analysis, and it completed as expected.
6/

Then I went into the /etc/systemd/system directory (mounted under /mnt). Sure enough 'tmp.mount' was symlink to /dev/null. That was the result of the 'systemd mask...' operation I had done earlier. I reset the symlink to its original value:

sudo ln -sf /usr/lib/systemd/system/tmp.mount .

and rebooted. Now my machine came up as usual. Whew!
5/

After years of experience trusting Debian, I merrily ran that command and rebooted, only to find that the system failed to boot. Some other services, that apparently depend on the tmpfs arrangement, failed to start, and froze the boot process!

Well, now I'm in a pickle. After rummaging thru my brain to decide how to solve this, out comes the USB flash disk with my Debian ISO for installing. I booted into the Live system, mounted the root file system under /mnt.
4/

#gisChat #RStats
For usual day-to-day use that's fine. And it all disappears at each reboot. So for most applications the system works fine.

However I'm running a spatial analysis process with somewhat large raster layers. Due to this tmpfs limitation, the /tmp ramdisk filled up before completing, and the process failed. (I had to manually delete the stale files in /tmp/Rtmpxxxxxx in order to continue working). So I clearly need more /tmp space to run this analysis.
2/

#gisChat #rstats

I fell into a bit of a rabbit hole, using the R package {terra} on Debian 13.
Not sure where to post this so I'm just throwing it out here. (TLDR, long post coming...)

Debian now has defined '/tmp' as a ramdisk rather than an actual file system directory. I guess this has some speed advantages, and saves writes and deletes on the system disk. But it's limited to 50% of the computer RAM. So if I have 16 GB in my machine, my "/tmp" files cannot be more than 8 GB.

1/

RE: @fionag11@mapstodon.space

If we just renamed to "orientation arrow" would that solve the problem?
The issue is with the 99% who insist N is at the top, not the addition of cartographic elements.
:thinking: