Re: Shared memory - using ps(1)

[Bill Moseley <moseley@hank.org>]

On Tue, May 02, 2006 at 01:42:28AM -0400, Roberto C. Sanchez wrote:
> I get your point.  Hopefully you get mine :-)

I hope I get both points. ;)

My 20 processes of statically linked MySQL may share memory between
themselves (because they were forked from a parent??) but if MySQL
was linked with libc, that won't be shared with other dynamically
linked programs that use libc.  So libc is using double the memory
in this case.



I was looking at /proc/<pid>/smaps for Apache2, postgresql 8.1, and
(the famous statically linked MySQL).  I just wrote a little perl
script to tally up the various counts -- which is probably
meaningless (since you can't add up a bunch of processes' shared
memory because it's, well shared).

The program displays the /proc/<pid>/smaps info for all processes
padded and then shows totals.  I've clipped all output except for two
pids and the total.


$ perl smaps.pl `pidof apache2`  (5 pids)
[...]

[9677]
           size:    62028K
            rss:    43140K
   shared_clean:     4388K
   shared_dirty:    18500K
  private_clean:        0K
  private_dirty:    20252K

[9675]
           size:    56360K
            rss:    38164K
   shared_clean:     4472K
   shared_dirty:    21144K
  private_clean:        0K
  private_dirty:    12548K

[totals]
           size:   299780K
            rss:   206820K
   shared_clean:    20860K
   shared_dirty:    97476K
  private_clean:        0K
  private_dirty:    88484K

So, if all those processes are in physical RAM at the same time can I
assume that they use at least 88484K (plus whatever is shared in all)?

Now looking at MySQL:

$ perl smaps.pl `pidof mysqld--kernis`  (21 pids)
[...]

[9265]
           size:   145668K
            rss:    36576K
   shared_clean:     4040K
   shared_dirty:        0K
  private_clean:        0K
  private_dirty:    32536K

[9264]
           size:   145668K
            rss:    36576K
   shared_clean:     4040K
   shared_dirty:        0K
  private_clean:        0K
  private_dirty:    32536K

[totals]
           size:  3059028K
            rss:   768096K
   shared_clean:    84840K
   shared_dirty:        0K
  private_clean:        0K
  private_dirty:   683256K

Here's where I'm curious. 680MB of private memory.  It's likely that
not all of that will be in RAM at the same time.  (For one thing, the
machine does not have that much traffic).  But, in general, as
connections are made to MySQL don't they basially round-robin.  In
other words, say 10 of the 20 processes have much of their private
memory swapped out.  It's not like only the 10 with all their memory
still in physical RAM will somehow handle all connections to MySQL.

So my guess is:

Even with low MySQL traffic having more processed than needed
causes swapping to happen since connections to MySQL will cycle
through all MySQL processes and potentially bring them in and out of
physical RAM.

Of course, it's not that simple. ;)







-- 
Bill Moseley
moseley@hank.org