making editors/openoffice-devel within poudriere: out of swap with 10 GByte

[Matthias Apitz]

El día Monday, February 08, 2016 a las 04:08:50PM -0500, Allan Jude escribió:

> On 2016-02-08 15:38, Matthias Apitz wrote:
> >
> > Hello,
> >
> > I have a poudriere oven, amd64 based on r292778 and poudriere 3.1.99.20151204
> > which served fine to build some 1700 ports. Yesterday I started to make
> > in addition editors/openoffice-devel, and it ended up eating all
> > attached swap space and at the end today with out of swap after some ~15 hours...
> >
> > The swap is 4 GByte /dev/ada0p3 and in addition 6 plain files of 1 GByte added
> > as swap devices.
> >
> > What I do not understand, while watching the situation with top, there
> > have only be running as maximum 2 processes clang++, and always starting new
> > processes for new foo.cxx files, and theses have had a memory of some
> > let's says 200 Mbytes each, as max. How can this lead to eat up 10 GByte swap
> > space?
> >
> > 	matthias
> >
> 
> Where you using the tmpfs feature of poudriere? I believe it is on by 
> default. This will obviously use up a bunch of ram and cause swapping.

Thanks. I commented out

# USE_TMPFS=

and now the picture in top is as expected. Used swap isn't growing after
2++ hours:

last pid: 39826;  load averages:  1.11,  1.08,  1.03 up 0+02:20:00  15:38:12
41 processes:  2 running, 39 sleeping
CPU: 47.0% user,  0.0% nice,  4.7% system,  0.0% interrupt, 48.2% idle
Mem: 106M Active, 2867M Inact, 679M Wired, 340M Buf, 220M Free
Swap: 3979M Total, 11M Used, 3968M Free

  PID USERNAME    THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
39773 root          1  78    0   128M   107M CPU0    0   0:02  98.28% clang++
  867 guru          1  20    0 82872K  3772K select  1   0:02   0.01% sshd
32359 root          1  20    0 20096K  3300K CPU1    1   0:00   0.04% top
 1070 root          1  52    0  8404K  2848K wait    0   0:26   0.35% sh
  296 root          1  20    0 20472K  2108K select  1   0:00   0.00% wpa_supplicant
39826 root          1  52    0  6220K  1424K nanslp  0   0:00   0.04% sleep
  752 root          1  20    0 10412K  1228K select  1   0:00   0.00% powerd

Watching with tail -f the log file, it seems even, that clang++ is compiling 
faster than before.

	matthias
-- 
Matthias Apitz, ✉ guru at unixarea.de, ⌂ http://www.unixarea.de/  ☎ +49-176-38902045
UNIX since V7 on PDP-11 | UNIX on mainframe since ESER 1055 (IBM /370)
UNIX on x86 since SVR4.2 UnixWare 2.1.2 | FreeBSD since 2.2.5