Crash problem..

["George Henry C. Daswani" <gdaswani@odc.net>]

While stress testing 2.0.7 under RedHat Linux 7.1, Berkeley DB 2.7.7 @
480,000 directory entries.  2 CPU Dell's with 1 MB of RAM, kernel 2.4.2.


Apr 24 15:40:35 oakley slapd[11188]: <= key_change 0
Apr 24 15:40:35 oakley slapd[11188]: => key_change(DELETE,1f209)
Apr 24 15:40:35 oakley slapd[11188]: <= key_change 0
Apr 24 15:40:35 oakley slapd[11188]: => key_change(DELETE,1f209)
Apr 24 15:40:35 oakley slapd[11188]: ch_calloc of 2748252165 elems of 4 bytes fa
iled

This seems to be a threading issues, especially under updates..  It seems
that running one writer doesn't cause the problem, but having
simultaneous writers shuts down the ldap server within 5 minutes of
operation.

the dbcacachesize and cachesize is set to default, so that shouldn't be a
problem.

It seems that ch_alloc is trying to allocate *A LOT OF MEMORY* there and
it doesn't seem right..  Looks very much like stack corruption.
Allocating around 10 GB of mem, off course it will fail.

George