[Samba] Slowness problems with Samba 3.0.1, Solaris, and Clearcase

[Kost, Kathy]

Hello --

I recently installed Samba 3.0.1 on our Solaris 8 servers (as well as
our RedHat servers) and I wanted to see if I could get some feedback on
a few things.

We are using Samba in conjunction with Clearcase for our software 
developers and because of that, we have oplocks turned off (also level2
oplocks = no as well).  We are authenticating to our ADS server, but 
joining the domain with security=domain.

We are seeing that initial authentication tends to be slow, and we
see log files of the sort of log.38.255.38.1 being created as it does 
initial setup.  Then after authenticating, these logs are closed and
we start seeing logs based on the hostname of the pc (i.e. log.kkostpc).
I had not noticed this type of logging before we started using 
security=domain.  In the past we used security=server, but had problems 
and have since noticed the docs warning against using this.  Is this 
typical behavor when initially authenticating a connection?  I'm concerned 
about why the initial authentication slowness.  Also, one of our developers saw
a huge increase in the amount of time to build a dynamic Clearcase view (12 hours
instead of the usual 6 hours).  This may be an unrelated problem, but I thought I'd
mention it.

I'm also seeing the following errors and would like to know some info
about them.  THe first is:

[2004/01/03 10:49:26, 0] lib/util_sock.c:set_socket_options(160)
  Failed to set socket option SO_KEEPALIVE (Error Invalid argument)

I noticed that SO_KEEPALIVE is compiled by default into Samba, and on
Solaris it is on by default.  The value of tcp_keepalive_interval is
7200000.  I can change this value, but don't want to mess with it unless
I'm really sure of what I'm doing.  Why the invalid argument error?

Next one is this:

[2004/01/10 00:06:12, 0] locking/posix.c:posix_fcntl_lock(657)
  posix_fcntl_lock: WARNING: lock request at offset 0, length 
922337203685477900
0 returned
[2004/01/10 00:06:12, 0] locking/posix.c:posix_fcntl_lock(658)
  an Invalid argument error. This can happen when using 64 bit lock 
offsets
[2004/01/10 00:06:12, 0] locking/posix.c:posix_fcntl_lock(659)
  on 32 bit NFS mounted file systems.
[2004/01/10 00:06:12, 0] locking/posix.c:posix_fcntl_lock(673)
  Count greater than 31 bits - retrying with 31 bit truncated length.

I got a lot more of these before I turned oplocks off.  My assumption
so far is that this is probably not a major problem and due to the client 
requesting an invalid lock request address.  I am not sharing out any NFS 
files systems, though, so I sometimes wonder where this is coming from.  I 
see in the docs that on unix, Samba assumes 32 bit lock offsets, even when 
compiled on a 64 bit machine.  Any more info on this would be helpful.

And finally, here is the last:

[2004/01/12 09:54:48, 0] lib/util_sock.c:set_socket_options(160)
  Failed to set socket option TCP_NODELAY (Error Invalid argument)
[2004/01/12 09:54:48, 0] lib/util_sock.c:get_peer_addr(948)
  getpeername failed. Error was Transport endpoint is not connected
[2004/01/12 10:16:45, 0] smbd/server.c:open_sockets_smbd(377)
  open_sockets_smbd: accept: Software caused connection abort

(Note:  I get the "getpeername failed" error also on RedHat, but not
the keepalive or tcpdelay errors)

Again, not sure why the failure to set this option.  Any clues on this
is appreciated.

Thanks for the help on any of these subjects.

Kathy