A NOTE has been added to this issue.
======================================================================
http://www.dbmail.org/mantis/view.php?id=462
======================================================================
Reported By: fehuq
Assigned To:
======================================================================
Project: DBMail
Issue ID: 462
Category: IMAP daemon
Reproducibility: always
Severity: major
Priority: normal
Status: new
target:
======================================================================
Date Submitted: 29-Nov-06 15:13 CET
Last Modified: 10-Dec-06 06:26 CET
======================================================================
Summary: imap daemon appears to leak memory on FETCH command
Description:
I delivered several thousand messages to an account on a test server, and
set up an IMAP account in Thunderbird to access it. Thunderbird ran its
junkmail controls on the folder, which issues a FETCH command for each
message. I noticed later that the dbmail-imapd child process handling my
connection had used almost all of the system's free memory and swap.
As best as I can tell, the process usually doesn't free memory used after
a message is fetched. I'm attaching a text file containing top's output of
the process at 3 second intervals. The memory usage does drop a bit in some
of the intervals, but overall it gets large fairly fast.
Killing, or having dbmail restart the child process after 1 connection
does successfully free the memory, though. The pop server does not show
the same behavior.
======================================================================
Relationships ID Summary
----------------------------------------------------------------------
has duplicate 0000469 dbmail-imapd fetches all available memory
======================================================================
----------------------------------------------------------------------
fehuq - 29-Nov-06 15:24
----------------------------------------------------------------------
I should have added about the top output that it's just a small sample,
over the course of about a minute and a half of a new process as I fetch
all the messages. It's not meant to show a process using all system
memory.
----------------------------------------------------------------------
aaron - 09-Dec-06 06:09
----------------------------------------------------------------------
Using the fetches you've shown and running through 'valgrind dbmail-imapd
-n' and typing them in by hand, against my testing database, I found no
active leaks.
Maybe the leaks depend upon the messages themselves? (*ugh*)
----------------------------------------------------------------------
Valen - 09-Dec-06 08:45
----------------------------------------------------------------------
I have a system that exhibits this symptom, I don't have the skill to run
the test but dbmail can be built on the system (and has been in the past)
if an admin wants to login and check it out let me know.
----------------------------------------------------------------------
paul - 09-Dec-06 18:46
----------------------------------------------------------------------
I've been busy running valgrind, and I can't find *any* significant leaks
either. But I am working on reducing memory usage in dbmail-message.c by
doing less copying which may affect this bug when I merge that change.
----------------------------------------------------------------------
Valen - 10-Dec-06 01:40
----------------------------------------------------------------------
attached files with memory graph from massif, this is the result of copying
100 emails with imapcopy
----------------------------------------------------------------------
Valen - 10-Dec-06 06:26
----------------------------------------------------------------------
found perhaps the causes of the problems and i got debug turned on so it
should be extra helpfull.
Called from:
41.8% : 0x4128031: g_malloc (in /usr/lib/libglib-2.0.so.0.1000.2)
24.8% : 0x401C3EB: posix_memalign (vg_replace_malloc.c:384)
15.4% : 0x4127F1E: g_realloc (in /usr/lib/libglib-2.0.so.0.1000.2)
11.0% : 0x4434D69: my_realloc (in /usr/lib/libmysqlclient.so.15.0.0)
1.8% : 0x4434B86: my_malloc (in /usr/lib/libmysqlclient.so.15.0.0)
1.8% : 0x443B415: my_once_alloc (in /usr/lib/libmysqlclient.so.15.0.0)
0.7% : 0x4127FB9: g_malloc0 (in /usr/lib/libglib-2.0.so.0.1000.2)
and 33 other insignificant places
== 1 ===========================
Context accounted for 41.8% of measured spacetime
0x4128031: g_malloc (in /usr/lib/libglib-2.0.so.0.1000.2)
Called from:
33.4% : 0x8052491: mopen (memblock.c:64)
19.8% : 0x805255E: __m_blkadd (memblock.c:435)
1.5% : 0x405ED0A: g_mime_filter_set_size (in
/usr/lib/libgmime-2.0.so.2.1.19)
and 40 other insignificant places
---------------------------------
Context accounted for 24.8% of measured spacetime
0x401C3EB: posix_memalign (vg_replace_malloc.c:384)
Called from:
24.8% : 0x4136098: (within /usr/lib/libglib-2.0.so.0.1000.2)
---------------------------------
Context accounted for 15.4% of measured spacetime
0x4127F1E: g_realloc (in /usr/lib/libglib-2.0.so.0.1000.2)
Called from:
14.2% : 0x4109AB0: (within /usr/lib/libglib-2.0.so.0.1000.2)
8.6% : 0x413A832: (within /usr/lib/libglib-2.0.so.0.1000.2)
and 20 other insignificant places
---------------------------------
Context accounted for 11.0% of measured spacetime
0x4434D69: my_realloc (in /usr/lib/libmysqlclient.so.15.0.0)
Called from:
11.0% : 0x445BF37: net_realloc (in /usr/lib/libmysqlclient.so.15.0.0)
---------------------------------
Issue History
Date Modified Username Field Change
======================================================================
29-Nov-06 15:13 fehuq New Issue
29-Nov-06 15:13 fehuq File Added: top_3sec_interval.txt
29-Nov-06 15:24 fehuq Note Added: 0001633
06-Dec-06 12:41 fehuq File Added: dbmail-imap-fetches.txt
09-Dec-06 05:51 aaron Relationship added has duplicate 0000469
09-Dec-06 06:09 aaron Note Added: 0001661
09-Dec-06 08:45 Valen Note Added: 0001662
09-Dec-06 18:46 paul Note Added: 0001663
10-Dec-06 01:39 Valen File Added: massif.13357.gif
10-Dec-06 01:39 Valen File Added: massif.13357.txt
10-Dec-06 01:40 Valen Note Added: 0001664
10-Dec-06 06:26 Valen Note Added: 0001665
======================================================================
