[Samba] Client Disconnects

[Nicholas McDowell]

Apologies for the duplication, this was sent before completion.

Hi,

I'm running samba 2.2.7 on redhat advanced server with approximately 40
win2k clients and having a serious problem with the clients.

Intermittently the clients will lose their connection to the server for
approximately 30 seconds.  It's like the socket connection is being reset or
something.

Here's a sample output from a client log file:

[2003/04/12 15:13:08, 0] lib/util_sock.c:get_socket_addr(1012)
  getpeername failed. Error was Transport endpoint is not connected
[2003/04/12 15:13:08, 0] lib/util_sock.c:write_socket_data(499)
  write_socket_data: write failure. Error = Connection reset by peer
[2003/04/12 15:13:08, 0] lib/util_sock.c:write_socket(523)
  write_socket: Error writing 4 bytes to socket 5: ERRNO = Connection reset
by peer
[2003/04/12 15:13:08, 0] lib/util_sock.c:send_smb(703)
  Error writing 4 bytes to client. -1. (Connection reset by peer)
[2003/04/12 15:13:09, 0] lib/util_sock.c:get_socket_addr(1012)
  getpeername failed. Error was Transport endpoint is not connected
[2003/04/12 15:13:09, 0] lib/util_sock.c:write_socket_data(499)
  write_socket_data: write failure. Error = Connection reset by peer
[2003/04/12 15:13:09, 0] lib/util_sock.c:write_socket(523)
  write_socket: Error writing 4 bytes to socket 5: ERRNO = Connection reset
by peer
[2003/04/12 15:13:09, 0] lib/util_sock.c:send_smb(703)
  Error writing 4 bytes to client. -1. (Connection reset by peer)


And another client at the same time:

[2003/04/12 15:13:08, 0] lib/util_sock.c:write_socket_data(499)
  write_socket_data: write failure. Error = Connection reset by peer
[2003/04/12 15:13:08, 0] lib/util_sock.c:write_socket(523)
  write_socket: Error writing 319 bytes to socket 12: ERRNO = Connection
reset by peer
[2003/04/12 15:13:08, 0] lib/util_sock.c:send_smb(703)
  Error writing 319 bytes to client. -1. (Connection reset by peer)


All the clients are using mapped drives which will also intermittently get
red crosses over the icon indicating loss of connection.  All clients are
involved in a render network and are accessing the server constantly so this
can't be a inactivity timeout, could it?

Some smd.conf parameters I've been playing with:

   oplocks = False
   level2oplocks = False
   debug level = 2 
   keepalive = 300
   
   socket options = TCP_NODELAY SO_KEEPALIVE SO_RCVBUF=8192 SO_SNDBUF=8192



Any help would be greatly appreciated.

Cheers
Nicholas