Printing causing failed connections

[Ken Hamer (B83 131)]

Dear Folks,

I've got a generally well-behaved Solaris 2.5.1 machine running a generally 
well-behaved version of 1.9.17p4 running as a daemon.  Everything works 
beautifully including browsing with the help of a central WINS database on an 
NT box and printing using a group of HP printers with JetDirect cards.

BUT...... After printing a document, Samba appears to hang up the connection, 
although another seems to start pretty quickly.  Running netstat before 
printing produces something like the following:

server.netbios-   client.1065            8557      0  8760      0 ESTABLISHED

Seconds after spooling the print job, the following appears:

server.netbios-   client.1065            8557      0  8760      0 CLOSE_WAIT

The connection will stay in this state until the user tries to access 
something on the network drive or until it times out.  smbstatus reflects the 
above and shows a connection until time-out.

Normally, when the user tries to access anything on the network drive, the 
smbd daemon, as expected, kicks off a new process, a new connection is 
established and everybody is happy.

HOWEVER, some clients drop the network drive and their users find that they 
must re-connect.  Sometime this is not possible until the old connection times 
out.  Both NT and Windows95 clients are affected.  I SUSPECT that this is more 
of a client problem than a Samba/UNIX one but I can't prove it.

BTW, the print job which causes the problem always comes out OK.

Has anyone seen this phenomenon before and if so, has anyone got a 
fix/workaround for it?  Any information would be appreciated.

Ken Hamer

khamer at axion.bt.co.uk