I’m joining the club of people complaining about /tmp on tmpfs

Let me start this post by saying that most of the time, I agree with Lennart’s reasoning about technical decisions, and I do believe he is doing a great job for our community with his work on PulseAudio/systemd/the journal. That said…

The tmp on tmpfs proposal, starting with Fedora 18, is a big mistake. Rich Jones has defended this position quite well, so let me just quote him in what I believe it is the central problem with this proposal:

Everyone must now be careful never to store a file in /tmp that might grow large, nor too many files at the same time. Every last little utility must be checked, including ones that aren’t part of your distro. Every user must be “re-educated” not to use /tmp for temporary files.

Having to bring that upon users is an unnecessary mess, for questionable advantages. No, I don’t consider ‘performance improvements’ running bonnie++ on tmpfs something that’s even remotely interesting (and by the way, of course there’s a performance improvement, it’s running in the goddamn RAM). You may not be able to change every single utility, since some of them might be proprietary.

There’s a single valid point about this proposition, which is to spare SSDs from some extra write cycles, possibly expanding the lifespan of the storage, but if we weight the pros and cons, the cons win easily. Even the fact the default can be easily changed does not justify having to bring that upon the users.

So, as Rich, I hope that FESCO reconsiders this position, as for instance, Debian decided to revert it.

Published by lmr

I'm yet another Software Engineer working to improve to Open Source Software (OSS) Testing Stack. About me

4 thoughts on “I’m joining the club of people complaining about /tmp on tmpfs

  1. I wasn’t personally aware of the proposal until now, but this also seems like a noteworthy con:
    “This may also impact analysis of system crashes as /tmp will be lost on a non-clean shutdown as well as an orderly one. “

  2. The ‘solution’ for this is to change crash analysis programs to write to /var/tmp. Again, lots of patches for not a lot of benefits.

  3. Now having read through the FESCO ticket…
    I agree with a couple of points… even if there are performance improvements with /tmp on tmpfs, if programs switch to using /var/tmp (which is on disk), then the performance improvement is negated!

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: