[Samba] 100% CPU eaten -- tdb_fetch failed

[Fabien Chevalier]

Hello,

From time to time Samba 3.0.2 performance suddenly fails bellow
what is acceptable.
One smbd process eats between 30% and 100% of CPU usage, and
for the machine associated with the smbd process i got
dozens of (debug level 2):

[2004/02/25 17:59:18, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:19, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:20, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:20, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:20, 0] smbd/connection.c:register_message_flags(220)
  register_message_flags: tdb_fetch failed
[2004/02/25 17:59:20, 0] smbd/connection.c:register_message_flags(220)

I tried to stop and restart Samba without success.
The client machine recreates the connection, Samba forks a new smbd,
which sill eats too much CPU.

I looked at previous posts, but didn't found any explanation regarding
tdb_fetch failed issues.

How can i know which tdb is faulty?
Is there another way to solve this issue than removing the tdb ?
(...but for now i don't know which one it is...)

Bye,

Fabien