Solved! Re: [ARGUS] Running out of memory when doing a rasort

[eric]

On Fri, 2004-04-30 at 01:32:23 -0500, eric proclaimed...

> Ah yes, more fun problems.
> 
> During a large spike in traffic, we noticed that rasort dies off
> around 500MB being loaded into memory.
> 
> $ ls -l argus.2004.04.30.01.10.00.tmp
> -rw-r-----  1 me flows  329519468 Apr 30 01:10 argus.2004.04.30.01.10.00.tmp
> 
> Turns up with...
> 
> rasort[99756]: RaCopyArgusRecord(0xbfbfe990) ArgusCalloc(1, 148) error Cannot
> allocate memory
> 

Ha! I solved my own problem.

After looking through the LINT kernel config, I found that I could
increase the hard limits of segment and stack sizes.

options MAXDSIZ="(1380*1024*1024)"      #added for rasort! 20040430
options DFLDSIZ="(1380*1024*1024)"      #added for rasort! 20040430
options MAXSSIZ="(1024*1024*1024)"      #added for rasort! 20040430

1024*1024*1024 == ~1GB

After this, rasort works just fine! Take a look...

  PID USERNAME PRI NICE  SIZE    RES STATE    TIME   WCPU    CPU COMMAND
42869 eric      52   2   634M   634M RUN      0:45 98.09% 88.28% rasort

Wow, thats a lot of memory. :)

Also, doing a ulimit -a shows this.....

data seg size           (kbytes, -d)  1413120
stack size              (kbytes, -s)  1048576

So I appear to be in good shape.

Hopefully if anyone is using FreeBSD and they run into this problem,
they can give it a try.

- Eric