Re: CURRENT: memory leak?

Am Fri, 29 Jul 2016 15:03:04 -0400
Allan Jude <allanjude_at_freebsd.org> schrieb:

> On 2016-07-29 14:04, O. Hartmann wrote:
> >
> > I realise an exorbitant memory usage of FreeBSD CURRENT ( FreeBSD 12.0-CURRENT #16
> > r303470: Fri Jul 29 05:58:42 CEST 2016 ). Swap space gets eaten up while building
> > world/kernel and/or ports very quickly.
> >
> > I see this phenomenon on different CURRENT systems with different RAM (but all ZFS!).
> > No box is less than 8 GB RAM: one 8GB, another 16, two 32 GB. An older XEON Core2Duo
> > server with postgresql 9.5/postgis acting on some OSM data etas up all of its 32 GB
> > and additional 48GB swap - never seen before with 11-CURRENT.
> >
> > I didn't investigate the problem so far since I realized this memory hunger of
> > 12-CURRENT just today on several boxes compiling world, eating up all the memory,
> > staring swapping and never relax even after hours from the swapped memory.
> >
> > Is this a known phenomenon or am I seeing something mystique?
> >
> > Regards,
> >
> > Oliver
> >  
> 
> Do you have the output of 'top', the first few lines
> 
> Specifically, is there very high 'Other' usage, on the ZFS ARC line?
> 

I had to reboot the worst box since it didn't answer on any input or ssh. I also just
restarted the other systems after buildworld/buildkernel of the most recent src ...

I try VBox - it complained last time not having enough memory after I simply
buildworld/portupdated ...

This is head of top just before I start buildworld/buildkernel on the 8 GB (dual core/
4 threads box):

last pid:  1645;  load averages:  0.13,  0.15,
0.28
up 0+01:08:06  21:24:01 61 processes:  1 running, 60 sleeping CPU:  0.2% user,  0.0%
nice,  0.4% system,  0.0% interrupt, 99.4% idle Mem: 118M Active, 547M Inact, 4376M
Wired, 667M Buf, 2788M Free ARC: 2411M Total, 836M MFU, 363M MRU, 32K Anon, 27M Header,
1185M Other Swap: 64G Total, 64G Free

  PID USERNAME       THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
 1002 root             1  20    0 12434M 49192K select  2   0:39   1.34% Xorg
  514 root             1  20    0 12748K  2252K select  0   0:22   0.27% moused
 1027 ohartmann        1  20    0   125M 20488K select  0   0:02   0.14% wmaker
 1634 ohartmann        1  20    0 76020K 11988K select  2   0:00   0.09% xterm
 1645 root             1  20    0 22260K  3588K CPU1    1   0:00   0.08% top
  403 root             1  20    0  9588K  5060K select  2   0:00   0.01% devd
  900 root             1  20    0 88148K  9768K select  2   0:00   0.00% sendmail
  563 root             1  20    0 12608K  2476K select  0   0:00   0.00% syslogd
  793 root             1  20    0 22764K 12632K select  0   0:00   0.00% ntpd
[...]

This is the top after "make -j4 buildkernel buildworld" (filemon.ko enabled) and
afterwards a "make update" in /usr/ports which resides on an ZFS 3TB disk and is svn
backed and after that 5 minutes wating/relaxing:

last pid: 11935;  load averages:  0.10,  0.55,
0.64
up 0+01:21:10  21:37:05
66 processes:  1 running, 65 sleeping
CPU:  0.4% user,  0.0% nice,  0.6% system,  0.0% interrupt, 99.0% idle
Mem: 308M Active, 2805M Inact, 4431M Wired, 775M Buf, 284M Free
ARC: 2470M Total, 1007M MFU, 279M MRU, 32K Anon, 31M Header, 1153M Other
Swap: 64G Total, 64G Free

  PID USERNAME       THR PRI NICE   SIZE    RES STATE   C   TIME    WCPU COMMAND
 1002 root             1  21    0 12433M 44044K select  1   1:02   1.47% Xorg
  514 root             1  20    0 12748K  2248K select  1   0:24   0.29% moused
 1027 ohartmann        1  21    0   125M 16536K select  1   0:09   0.18% wmaker
 1645 root             1  20    0 22260K  4096K CPU2    2   0:01   0.08% top
 1032 ohartmann        1  20    0 86260K 20344K select  3   0:06   0.06% xterm
 1634 ohartmann        1  20    0 76020K 10304K select  1   0:01   0.03% xterm
21103 ohartmann        1  20    0 76020K 10296K select  1   0:00   0.02% xterm
21553 ohartmann        1  20    0 26776K  6136K select  2   0:00   0.00% ssh
  793 root             1  20    0 22764K 12632K select  1   0:00   0.00% ntpd
  820 root             1  20    0 43744K  3184K select  0   0:00   0.00% saned
 1026 ohartmann        1  20    0 33592K  4028K select  0   0:00   0.00% gpg-agent
[...]


I do not understand much of the memory partitioning, but starting with ~ 2 GB free RAM
and ending up with < 300MB free RAM - if correctly interpreted - seems weird.

I'm out of the office, so I can not access the box that is literaly "exploding" right now.

Hope you can see something in the glass sphere ...

Kind regards, thank you very much in advance,

Oliver