- start SSSD (systemctl start sssd) with normal sssd.conf
- call:
    gdb -ex "call teardown_watchdog()" -ex 'call talloc_enable_null_tracking()'  -ex 'call talloc_report_full(0, debug_file)'  -ex 'detach' -ex 'quit' -p $(pidof sssd_be)
   and
   gdb -ex "call teardown_watchdog()" -ex 'call talloc_enable_null_tracking()'  -ex 'call talloc_report_full(0, debug_file)'  -ex 'detach' -ex 'quit' -p $(pidof sssd_nss)
- wait until the sssd_be or sssd_nss process use lots of memory and don't give it back
- call:
    gdb -ex "call teardown_watchdog()" -ex 'call talloc_enable_null_tracking()'  -ex 'call talloc_report_full(0, debug_file)'  -ex 'detach' -ex 'quit' -p $(pidof sssd_be)
   or
   gdb -ex "call teardown_watchdog()" -ex 'call talloc_enable_null_tracking()'  -ex 'call talloc_report_full(0, debug_file)'  -ex 'detach' -ex 'quit' -p $(pidof sssd_nss)
   as root from the shell
- save the domain log and sssd_nss.log and restart SSSD to release the memory.