[Samba] could not login to samba server

[Yuval Hager]

Hi,

I am working with samba-3.0alpha20, and suddenly, one of the users could
not log in to the system. I worked with `wbinfo' and `id' (I am working
as security=daemon, using winbind in nsswitch.conf).

Here is what failed for `us2':
# id NTDOM\\us2
uid=10003 gid=1(NTDOM\Domain Users) groups=1(NTDOM\Domain Users)
/*== pay attention that there is no username in parentheses after the
uid */
# wbinfo -s S-1-5-21-1789293116-3789330910-3885930141-1150  
Could not lookup sid S-1-5-21-1789293116-3789330910-3885930141-1150

And here is what worked for `us3':
# id NTDOM\\us3
uid=1(NTDOM\us3) gid=1(NTDOM\Domain Users)
groups=1(NTDOM\Domain Users) # wbinfo -s
S-1-5-21-1789293116-3789330910-3885930141-1151
NTDOM\us3 1

All other wbinfo conversions (name to sid, sid to uid etc.) worked fine.
I was also able to list all domain users (wbinfo -u).

What I saw in the logs (log.winxp) was the following line:

[2003/07/13 12:21:19, 0]
auth/auth_util.c:get_user_groups_from_local_sam(670)
  Couldn't find passdb structure for UID = 10003 ! Aborting.

After 80 minutes, the issue was resolved by itself (I have `winbind
cache time = 15').

Any ideas?

Thanks,

--yuval
--
To unsubscribe from this list go to the following URL and read the
instructions:  http://lists.samba.org/mailman/listinfo/samba

Windows client disconnect identification

[Yuval Hager]

Hi,

The scenario I am testing is this:

- Open an MS word file from windows client A
- Disconnect the network cable from that windows client (emulate client
failure)
- Try to open the file from windows client B - file is locked.
- Wait a few minutes (see below - the exact time is the issue here)
- Try to open the file from windows client B - file is opened RW.

I am wondering what sets the time that it will take for samba to find
out that the first client is gone.

At first, I tried to use the deadtime variable - but it is only valid
when there are no open files (by the way, why?).

Then I tried to set various timeouts in the TCP socket definition
(socket options, various sysctl values), nothing worked.

The lock release time is incosistent, and goes between 10 to over 20
minutes. I am wondering how can I shorten this timeout..

Here a transcript from the log, this time shows a 23 minutes timeout:

[2003/02/28 18:22:42, 0] locking/posix.c:posix_fcntl_lock(662)
  on 32 bit NFS mounted file systems.
[2003/02/28 18:46:05, 0] lib/util_sock.c:read_socket_data(431)
  read_socket_data: recv failure for 4. Error = No route to host

Pay attention that the error came from read() and not from select(). How
can read timeout be controlled? Why does receive_message_or_smb() calls
receive_smb() without any timeout?

I am using samba-3.0-alpha20 on a Linux 2.4.18 machine, and windows 2k
clients.

Thanks,

--yuval