memory leaks in -current / Stoyan Zhekov <stoyan@silveregg.co.jp>

Hi, we have a problems with -current version. It's leaking.
The result of 'top' after about a week of work:

    15029 nobody     9   0  159m 159m 1268 S  0.6 15.8   0:00.02 pound

/etc/pound.cfg:

    ListenHTTPS 192.168.0.111,443 /etc/ssl/pound.pem

    UrlGroup ".*"
    BackEnd 192.168.0.111,80,1
    EndGroup

    User nobody
    Group nobody

Re: memory leaks in -current
Robert Segall <roseg(at)apsis.ch>

2004-04-27 12:36:05

[ FULL ]

On Tue, 2004-04-27 at 03:13, Stoyan Zhekov wrote:[...]

That doesn't prove much, does it? Depending on the system you run on and
the number of concurrent requests Pound may need a lot of memory.

If you really think there is a leak help us by at least pointing at some
likely candidates.[...]

Re: memory leaks in -current
Stoyan Zhekov <stoyan(at)silveregg.co.jp>

2004-04-28 02:23:04

[ FULL ]

>That doesn't prove much, does it? Depending on the system you run on and
 >the number of concurrent requests Pound may need a lot of memory.

The problem is now how much memory it use, but that used memory growing
constantly. On previous run it reached 254Mb before we to restart the
pound.

Re: memory leaks in -current
"Simon Matter" <simon.matter(at)ch.sauter-bc.com>

2004-04-28 08:19:41

[ FULL ]

>  >That doesn't prove much, does it? Depending on the system you run on
and[...]

I can confirm the problem. After running for about one month, pound has
grown to almost 600Mb which looks like being way too much. This is from
top:

  PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME COMMAND
11035 nobody    17   0  566M 364M   704 S    37.2 72.5 595:41 pound
28834 nobody    11   0  1524 1456  1172 S     0.5  0.2   0:00 perdition.pop3s
  658 root       9   0   280  232   192 S     0.1  0.0  11:40 syslogd
28899 nobody    12   0  1332 1256  1000 S     0.1  0.2   0:00 perdition.imaps

The box has 512M ram. Pound becomes quite slow until I restart it. This is
on RedHat 7.3.

Simon

Re: memory leaks in -current
"Simon Matter" <simon.matter(at)ch.sauter-bc.com>

2004-04-28 08:24:27

[ FULL ]

>>  >That doesn't prove much, does it? Depending on the system you run
on[...][...]

Sorry, forgot to mention, this is pound 1.7 but it was the same with older
releases IIRC.

Re: memory leaks in -current
Robert Segall <roseg(at)apsis.ch>

2004-04-28 13:53:36

[ FULL ]

On Wed, 2004-04-28 at 08:24, Simon Matter wrote:[...]

OK, thanks for the info. Any idea where this might be coming from?

Anybody else on the list observed same and/or has any suggestions?[...]

RE: memory leaks in -current
John D <jwdavid(at)ibizvision.com>

2004-04-28 16:37:24

[ FULL ]

We are using Pound-current (the one from 2004-04-05, there appears to be one
more recent then that [from the 26'th], but I don't recal seeing a message
about it). It looks like this after a restart:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND           
                                    31319 nobody     9   0  2392 2392 1664 S 
0.0  0.3   0:00.00 pound                                                  31320
nobody     1   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31324 nobody     9   0  2392 2392 1664 S  0.0 
0.3   0:00.00 pound                                                  31325
nobody     9   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31326 nobody     9   0  2392 2392 1664 S  0.0 
0.3   0:00.01 pound                                                  31352
nobody    11   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31359 nobody    10   0  2392 2392 1664 S  0.0 
0.3   0:00.00 pound

And this after a couple of days:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND           
                                    25991 nobody     9   0 29664  28m 1632 S 
0.0  3.3   0:02.71 pound                                                  25992
nobody    10   0 29664  28m 1632 S  0.0  3.3   0:04.33 pound                   
                              25996 nobody     9   0 29664  28m 1632 S  0.0 
3.3   0:00.00 pound                                                  25997
nobody     9   0 29664  28m 1632 S  0.0  3.3   0:00.00 pound                   
                              25998 nobody     9   0 29664  28m 1632 S  0.0 
3.3   0:00.01 pound                                                  31154
nobody    12   0 29664  28m 1632 S  0.0  3.3   0:00.00 pound                   
                              31158 nobody    16   0 29664  28m 1632 S  0.0 
3.3   0:00.02 pound                                                  31161
nobody    10   0 29664  28m 1632 S  0.0  3.3   0:00.00 p
 ound

I don't know if this indicates a problem or not, but I thought I would share
it. We need to restart Pound fairly frequently so I can't provide a more long
term view.

I have noticed that Pound seems to randomly (by random I mean I don't see a
pattern yet) stop logging. After a restart it is fine. This has happened twice
over the past week.

John D.

********** Original Email *********
** To:   pound(at)apsis.ch
** From: Robert Segall <roseg(at)apsis.ch>
** Date: Wed, 28 Apr 2004 13:53:36 +0200
**********

On Wed, 2004-04-28 at 08:24, Simon Matter wrote:[...]

OK, thanks for the info. Any idea where this might be coming from?

Anybody else on the list observed same and/or has any suggestions?[...]

RE: memory leaks in -current
John D <jwdavid(at)ibizvision.com>

2004-04-28 17:02:16

[ FULL ]

Never mind about not seeing mention of a new version :) It was burried in
another message (SSL session id stuff).

John D.

********** Original Email *********
** To:   pound(at)apsis.ch, roseg(at)apsis.ch
** From: John D <jwdavid(at)ibizvision.com>
** Date: Wed, 28 Apr 2004 07:37:24
**********

We are using Pound-current (the one from 2004-04-05, there appears to be one
more recent then that [from the 26'th], but I don't recal seeing a message
about it). It looks like this after a restart:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND           
                                    31319 nobody     9   0  2392 2392 1664 S 
0.0  0.3   0:00.00 pound                                                  31320
nobody     1   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31324 nobody     9   0  2392 2392 1664 S  0.0 
0.3   0:00.00 pound                                                  31325
nobody     9   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31326 nobody     9   0  2392 2392 1664 S  0.0 
0.3   0:00.01 pound                                                  31352
nobody    11   0  2392 2392 1664 S  0.0  0.3   0:00.00 pound                   
                              31359 nobody    10   0  2392 2392 1664 S  0.0 
0.3   0:00.00 pound

And this after a couple of days:

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND           
                                    25991 nobody     9   0 29664  28m 1632 S 
0.0  3.3   0:02.71 pound                                                  25992
nobody    10   0 29664  28m 1632 S  0.0  3.3   0:04.33 pound                   
                              25996 nobody     9   0 29664  28m 1632 S  0.0 
3.3   0:00.00 pound                                                  25997
nobody     9   0 29664  28m 1632 S  0.0  3.3   0:00.00 pound                   
                              25998 nobody     9   0 29664  28m 1632 S  0.0 
3.3   0:00.01 pound                                                  31154
nobody    12   0 29664  28m 1632 S  0.0  3.3   0:00.00 pound                   
                              31158 nobody    16   0 29664  28m 1632 S  0.0 
3.3   0:00.02 pound                                                  31161
nobody    10   0 29664  28m 1632 S  0.0  3.3   0:00.00 p
 ound

I don't know if this indicates a problem or not, but I thought I would share
it. We need to restart Pound fairly frequently so I can't provide a more long
term view.

I have noticed that Pound seems to randomly (by random I mean I don't see a
pattern yet) stop logging. After a restart it is fine. This has happened twice
over the past week.

John D.

********** Original Email *********
** To:   pound(at)apsis.ch
** From: Robert Segall <roseg(at)apsis.ch>
** Date: Wed, 28 Apr 2004 13:53:36 +0200
**********

On Wed, 2004-04-28 at 08:24, Simon Matter wrote:[...]

OK, thanks for the info. Any idea where this might be coming from?

Anybody else on the list observed same and/or has any suggestions?[...]