Re: (ITS#3938) out of memory leak causes segfault

[hyc@symas.com]

quanah@stanford.edu wrote:
> Full_Name: Quanah Gibson-mount
> Version: 2.3.5
> OS: Solaris 8
> URL: ftp://ftp.openldap.org/incoming/
> Submission from: (NULL) (171.66.182.82)
>
>
> In testing a 2 million size database, I found that once I start querying
> sequentially through the entries, all memory on the system gets exhausted, and
> the program core dumps.  Unfortunately, I cannot GDB the process as there is no
> memory available.  I'm hoping for a core file < 2GB that I can gdb, but no luck
> so far.
The valgrind output indicates there is no leak. Further investigation 
shows this is just a case of configuring too much cache for the 
machine's address space. The server in question had an entry cache of 2 
million and IDL cache of 6 million. At an average of 1KB per entry, a 
2Mentry cache will consume 2GB of memory. Most likely the process' 
address space was exhausted by the excessive cache settings.

This implies that the practical upper bound for the entry cache setting 
on a 32 bit machine is around ~1 million entries.

-- 
  -- Howard Chu
  Chief Architect, Symas Corp.  http://www.symas.com
  Director, Highland Sun        http://highlandsun.com/hyc
  OpenLDAP Core Team            http://www.openldap.org/project/