guix-daemon tmpfs out of space on parabola

  • Done
  • quality assurance status badge
Details
3 participants
  • Ludovic Courtès
  • Tobias Geerinckx-Rice
  • swedebugia
Owner
unassigned
Submitted by
swedebugia
Severity
normal

Debbugs page

S
S
swedebugia wrote on 6 Mar 2019 06:04
(name . Report new Guix bug)(address . bug-guix@gnu.org)
22f759c1-a41c-d3b8-08e7-fe4cda616894@riseup.net
Hi

I'm trying to build New Moon (unbranded palemoon browser) in parabola.
The tmpfs i 2G which is way too little it seems.

I read the man page of guix-daemon and the manual and could not find a
way to point guix-daemon to another tmpdir than /tmp.

Is this a bug?

In case it is not possible to change on the guix side (before we rewrite
the daemon) I think we should add a section about how to work around it
to the manual.

--
Cheers Swedebugia
Attachment: signature.asc
T
T
Tobias Geerinckx-Rice wrote on 6 Mar 2019 06:31
(name . swedebugia)(address . swedebugia@riseup.net)(address . 34768@debbugs.gnu.org)
8736o08494.fsf@nckx
Hullo,

swedebugia wrote:
Toggle quote (8 lines)
> I'm trying to build New Moon (unbranded palemoon browser) in
> parabola.
> The tmpfs i 2G which is way too little it seems.
>
> I read the man page of guix-daemon and the manual and could not
> find a
> way to point guix-daemon to another tmpdir than /tmp.

Well, TMPDIR= should Just Work :-)

You'll have to set it to the environment of guix-daemon (and
restart the daemon) for it to have any effect. How that's done
depends on your service manager/init system.

I guess a ‘set-tmpdir’ RPC could be added to the protocol. I
haven't thought through the security implications, and IMO it's
just papering over the real bug, which is…

Toggle quote (2 lines)
> Is this a bug?

Not sure what exactly you're referring to, but yes, I do think
that using /tmp instead of /var/tmp (or any non-tmpfs) is a bug by
modern(?) conventions. It's guaranteed to break eventually on
almost every non-Guix System, and when it doesn't it's an abuse of
RAM and swap space.

Kind regards,

T G-R
S
S
swedebugia wrote on 7 Mar 2019 06:20
(name . Tobias Geerinckx-Rice)(address . somebody@not-sent-or-endorsed-by.tobias.gr)(address . 34768@debbugs.gnu.org)
0E607D15-3AB6-410C-A609-AEB092844EF6@riseup.net
Tobias Geerinckx-Rice <somebody@not-sent-or-endorsed-by.tobias.gr> skrev: (6 mars 2019 15:31:13 CET)
Toggle quote (33 lines)
>Hullo,
>
>swedebugia wrote:
>> I'm trying to build New Moon (unbranded palemoon browser) in
>> parabola.
>> The tmpfs i 2G which is way too little it seems.
>>
>> I read the man page of guix-daemon and the manual and could not
>> find a
>> way to point guix-daemon to another tmpdir than /tmp.
>
>Well, TMPDIR= should Just Work :-)
>
>You'll have to set it to the environment of guix-daemon (and
>restart the daemon) for it to have any effect. How that's done
>depends on your service manager/init system.
>
>I guess a ‘set-tmpdir’ RPC could be added to the protocol. I
>haven't thought through the security implications, and IMO it's
>just papering over the real bug, which is…
>
>> Is this a bug?
>
>Not sure what exactly you're referring to, but yes, I do think
>that using /tmp instead of /var/tmp (or any non-tmpfs) is a bug by
>modern(?) conventions. It's guaranteed to break eventually on
>almost every non-Guix System, and when it doesn't it's an abuse of
>RAM and swap space.
>
>Kind regards,
>
>T G-R

Hi 😃

Thanks for the details. I agree that it is a bug.

I solved it by masking tmp.mount in systemd according to the arch wiki and thus disabled tmpfs for the whole thing.
--
Sent from my k-9 mail for Android.
Attachment: file
L
L
Ludovic Courtès wrote on 10 Mar 2019 10:29
(name . swedebugia)(address . swedebugia@riseup.net)(address . 34768@debbugs.gnu.org)
874l8a3afo.fsf@gnu.org
Hi swedebugia,

swedebugia <swedebugia@riseup.net> skribis:

Toggle quote (3 lines)
> I read the man page of guix-daemon and the manual and could not find a
> way to point guix-daemon to another tmpdir than /tmp.

guix-daemon honors the TMPDIR environment variable, so you can run:

TMPDIR=/some/where/else guix-daemon …


HTH!

Ludo’.
L
L
Ludovic Courtès wrote on 10 Mar 2019 10:29
control message for bug #34768
(address . control@debbugs.gnu.org)
8736nu3afc.fsf@gnu.org
tags 34768 notabug
close 34768
?
Your comment

This issue is archived.

To comment on this conversation send an email to 34768@patchwise.org

To respond to this issue using the mumi CLI, first switch to it
mumi current 34768
Then, you may apply the latest patchset in this issue (with sign off)
mumi am -- -s
Or, compose a reply to this issue
mumi compose
Or, send patches to this issue
mumi send-email *.patch