[Samba] Problems with "nobody" processes in Samba 3.0.4

[Jason]

Hi

I am having the exact same problem.  Did anyone every solve it?  It occurs on
both of my samba servers, one of which is not being used.

Jason


----- original message -----

Hi,

Hopefully someone can help me with this because its driving me up the 
wall. I admin a Samba PDC which authenticates through an LDAP backend. 
Both the samba server and pam authenticate through the entries in the 
LDAP database.

I recently upgraded to 3.0.4 to combat the M$ hotfix that destroyed 
password changing. Since then things have been squiffy. All runs fine 
(apart from a grouping problem that I shall describe later) until a 
rogue samba thread appears which is owned by "nobody".

   PID USER     PRI  NI  SIZE  RSS SHARE STAT %CPU %MEM   TIME CPU COMMAND
  2639 root      20   0 13844 3832   804 S    40.8  1.5  4317m   0 slapd
  7889 nobody    15   0  1528  492   372 S     4.9  0.1  76:19   0 smbd

This particular top output is probably a bad example because the nightly 
backups are running at the same time. However the exhaustion of slapd as 
shown above occurs at the same time as this "nobody" thread appears. 
When the backup is not running the smbd thread usually hits about 40% 
CPU as well leading to a very congested fileserver. At this point the 
network slows to a crawl, killing these processes stops the slapd cpu 
usage but seems then to corrupt peoples smb sessions which seems to 
suggest the process is actually associated with a user.

In trying to track down this bug I've rearranged the entire ldap tree; 
we used to have an "ou=smb" tree for all samba classes and "ou=People" 
and "ou=Group" trees for all the posix classes. These have ow been 
rearranged so that ou=People,ou=Computers and ou=Group exist with both 
their posix and samba attributes in each respective tree.

I would really, really appriciate any help that you people can give. 
I've had success tracking down samba problems in the past but this one 
has me.