Mailinglist Archive: opensuse-factory (435 mails)

< Previous Next >
Re: [opensuse-factory] Proposal: /tmp as tmpfs
On Fri, 10 Jul 2020 14:26:01 +0200, Richard Brown <rbrown@xxxxxxx>
wrote:

On Fri, 2020-07-10 at 14:20 +0200, H.Merijn Brand wrote:
On Fri, 10 Jul 2020 14:10:30 +0200, Richard Brown <rbrown@xxxxxxx>
wrote:


Those applications are in serious breach of both the FHS
recommendations and POSIX requirements.

I would imagine they have a vested interest in fixing that
problem.

The example I can come up with is a production system where 4
parties are involved: end-user application, management-layer,
database layer, database server. Of those the management layer has
been declared out- of
service (dead, no maint) but the whole of the suite still depends on
that for at least two more years (some weird law that causes this
requirement). The owner of that product is unwilling to make any
change
to the product. No CVE fixes, not bug fixes, no feature request or
whatever, so the engineers responsible for the end-user application
and
the database layer both have to work around those issues.

Working with old propriatary software sometimes really takes away
the fun out of ICT :(

And you're running this on Tumbleweed?

No :)

Though Tumbleweed has never bees so stable for me as it has the past 6
months. I'd almost say it is more stable than an old 42.3

I'm pretty sure you're not expecting us to support such bad practice
indefinately, right?

If it would be an option, why not?

Let me expand on that a bit

I've updated a 10-year old Windows box for my daughter by replacing the
hardware with similar hardware I had on the attic (faster CPU mainly),
then I added 4 Gb of RAM and the system worked fine. Slooooooow, but
fine. Then I replaced the RAID disks (2 x WD320 Gb) with a single SSD
disk of 512Gb and the boot time reduced from a whopping 13 minutes to a
mere 15 seconds. Me happy, daughter happy. (lets shuv' aside the real
problems I faced with incompatible drivers that cost me another three
days. A system with no network is kinda useless these days)

Then I thought, well, if it is that much faster, why not replace my
2 Tb root HDD disk on openSUSE 15.1 with an SSD and see if it gets that
much faster too. Long story short: it took me a week to read all the
steps to take if you don't want to install a new system afresh and some
things in cloning *really* work different in Linux, so what I ended up
is a working system with a 2Tb SSD that holds the root FS (xfs) and the
old disk still installed for the UEFI partition and the swap partition
(and I got an on-line backup of the root partition for free). The
system is now noticeably faster and I learned a lot, but it was not the
ease I got on Windows.

There will be people trying to install new Operating Systems and/or new
software on (very) old hardware and most of the time, this just works
(unlike with Windows: do not try to install Windows 10 on a 10 year old
PC). It will make hobbyists more enthusiastic about Linux and its
community. They get old hardware and build it into a hobby-system with
18 Gb disks and 2 Gb of RAM and take it that it is slow, but it WORKS!

Not allowing some app to write to /tmp is a different thing than saying
that if you do it is gone when you reboot. I for one would accept that
/tmp is slower when used with a HDD (well, I now have 1.5Tb free to be
mounted for /tmp) and have enough to dump data there over a /tmp that
is a bit quicker but takes away my RAM. I have a choice at this point
in time to use a disk as /tmp and use my RAM for something else. I do
have more than enough disks stacked away that I could use, but my
memory slots are all used and I do not have a real need to trade that
for faster /tmp.

Short: I like to have that option for a long time forward

--
H.Merijn Brand http://tux.nl Perl Monger http://amsterdam.pm.org/
using perl5.00307 .. 5.31 porting perl5 on HP-UX, AIX, and Linux
https://useplaintext.email https://tux.nl http://www.test-smoke.org
http://qa.perl.org http://www.goldmark.org/jeff/stupid-disclaimers/
< Previous Next >
This Thread