samba - Dec 1999 - Broken Pipe Errors

Hi there everyone!

One of the users complained to me that yesterday, he suddenly couldn't
access the network. He was online the whole day, then suddenly at around
6:45pm he couldn't write. A restart of his workstation rendered him
completely disconnected from the Samba domain.

Server is running Samba 2.0.6 on a RedHat 6.0 machine with Linux kernel
2.2.13. Client is running Windows95 OSR2 with a Xircom Cardbus Ethernet II
NIC. The drivers used by the client are updated, and the machine is
connected directly to the ethernet switch to which the server is
connected.

I checked the logs and found a lot of these (that's ... 5 screens worth?
Didn't count anymore):

[1999/12/07 18:46:06, 0] lib/util_sock.c:write_socket_data(537)
 write_socket_data: write failure. Error = Broken pipe
[1999/12/07 18:46:06, 0] lib/util_sock.c:write_socket(563)
 write_socket: Error writing 4 bytes to socket 6: ERRNO = Broken pipe
[1999/12/07 18:46:06, 0] lib/util_sock.c:send_smb(751)
 Error writing 4 bytes to client. -1. Exiting

Would anyone know why these happen? I don't know where to start working on
solving my problem with this connectivity (I can't just tell this
particular user to live with the problem, he's the CEO, my boss,
hahaha!!!).

BTW, FWIW, I'm also the one who had the problem with my Samba not being
able to perform my socket options for some wierd reason, my logs for that
problem read:

[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option SO_KEEPALIVE (Error Bad file descriptor)
[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option TCP_NODELAY (Error Bad file descriptor)
[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option IPTOS_LOWDELAY (Error Bad file descriptor)
[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option SO_KEEPALIVE (Error Bad file descriptor)
[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option SO_SNDBUF (Error Bad file descriptor)
[1999/12/06 07:08:00, 0] lib/util_sock.c:set_socket_options(148)
  Failed to set socket option SO_RCVBUF (Error Bad file descriptor)

Also for whatever it's worth, I've recompiled my Samba, now using EGCS
1.1.2-12 (before I was using GCC 2.95.1 and thought maybe that was where
my problem lay). My optimization flags are: "-O3 -mpentiumpro
-march=pentiumpro -malign-loops=2 -malign-jumps=2 -malign-functions=2".
The server is on a Celeron 400 with 128MB of RAM. :-)

Please help!!! Solutions would be great, other people saying this is their
problem too would be great too (for notes comparison), and people with
explanations to these (personally) cryptic messages would be great also!
I'm no super-techie, so all help will be most appreciated!

BTW, if it's not too much to ask, please CC me in your reply at
<jijo@leather-collection.com> as I am not subscribed to the list. I will
just check it regularly via the web interface. :-)

Thanks again! :-)

   [        Jijo Sevilla         ]
   [ jijo@leather-collection.com ]

On 8 Dec 99, at 12:17, Federico Sevilla III wrote:
> [1999/12/07 18:46:06, 0] lib/util_sock.c:write_socket_data(537)
>  write_socket_data: write failure. Error = Broken pipe
> [1999/12/07 18:46:06, 0] lib/util_sock.c:write_socket(563)
>  write_socket: Error writing 4 bytes to socket 6: ERRNO = Broken pipe
> [1999/12/07 18:46:06, 0] lib/util_sock.c:send_smb(751)
>  Error writing 4 bytes to client. -1. Exiting
I seen these exact same errors on our Sparc Ultra10 running Solaris: 

SunOS snowmass 5.6 Generic_105181-16 sun4u sparc SUNW,UltraSPARC-
IIi-Engine

However, I was never able to figure out what caused it because at the same 
time, other services would die with similiar errors (yellow pages, for 
example).  The only way to fix it was to reboot the Sparc.

Lately, we haven't seen the errors and we did three things:
a)  Upgraded Samba to 2.0.6
b)  Applied a bunch of patches to the Solaris kernel
c)  Removed some hubs that individuals had added (which caused the the 
total amount of segments to be exceeded).

Good Luck,


Chance C. Geurin (chance@rtlogic.com)
RT Logic! (http://www.rtlogic.com)
Colorado Springs, Colorado

"The greatness of a Nation and it's moral progress can be judged by the
way that it's animals are treated." ~~Gandhi

We have find a solution.

All N?IP ans names  was not write in etc/host.

I have a WinNT PDC, and a Win2k SQL server. I am not sure which selection is
having a 'broken pipe' issue, but I have noticed in the few messages
referring to this error that quite alot seem to be pointing to Win2k. Anyone
actually get around this?
 
2000/11/27 16:50:30, 0] lib/util_sock.c:write_socket_data(537)
  write_socket_data: write failure. Error = Broken pipe
[2000/11/27 16:54:16, 0] lib/util_sock.c:write_socket_data(537)
  write_socket_data: write failure. Error = Broken pipe
[2000/11/27 16:56:30, 0] lib/util_sock.c:write_socket_data(537)
  write_socket_data: write failure. Error = Broken pipe

 
Fred King
Vice President, Internet Commerce Systems
 
-------------- next part --------------
HTML attachment scrubbed and removed