inetd and Solaris

Peter Polkinghorne Peter.Polkinghorne at brunel.ac.uk
Thu Oct 1 13:02:28 GMT 1998


Environment: Solaris 2.5.1 Samba 1.9.18p3

We run samba from inetd (no special reason - just how its been done in the 
past).  Then on one server we got masses of Samba daemons running (up to 40 as 
a time trying to run - load average 40+!), giving following complaints:


Sep 28 14:39:06 zeus.brunel.ac.uk smbd[10728]: standard input is not a socket, 
a
ssuming -D option
Sep 28 14:39:06 zeus.brunel.ac.uk smbd[10742]: bind failed on port 139 
socket_ad
dr=0.0.0.0 (Address already in use)

... the second error message is because inetd has the socket.  The problem 
went away when I killed & restarted inetd - thus allowing one of the many smbd 
to grab the socket as deamon (verified with lsof).

I speculate that the problem was due to the Solaris inetd property of limiting 
40 connection in 60 seconds to a service - not much if whole classroom of NT 
machines starts up.  I do not see why this should lead to lots of smbd's being 
started with stdin not a socket.

Does anyone have any further explanation?

Should Solaris users be discouraged from using inetd to start smbd?
BTW I am aware of the -r option to inetd.

-- 
-----------------------------------------------------------------------------
| Peter Polkinghorne, Computer Centre, Brunel University, Uxbridge, UB8 3PH,|
| Peter.Polkinghorne at brunel.ac.uk   +44 1895 274000 x2561       UK          |
-----------------------------------------------------------------------------




More information about the samba mailing list