Samba error messages

[Jens Ebert]

Runnung Samba 2.0.6 on UX 11.0 we see various error messages in the syslog. Any clue what is wrong?

We are running the Samba-package that came with HP-UX. On other Samba-system we use the UX ports from the net without these errors.

Jan 29 10:48:45 server3 smbd[1107]: [2001/01/29 10:48:45, 0] lib/util_sock.c:(14
8)

Jan 29 10:48:45 server3 smbd[1107]: Failed to set socket option SO_KEEPALIVE (
Error Bad file number)

Jan 29 10:48:45 server3 smbd[1107]: Failed to set socket option TCP_NODELAY (E
rror Bad file number)

Feb 15 12:07:26 server3 smbd[23328]: [2001/02/15 12:07:26, 0] lib/util_sock.c:(4
74)

Feb 15 12:07:26 server3 smbd[23328]: read_socket_data: recv failure for 4. Err
or = Connection timed out

Feb 21 16:26:40 server3 smbd[1409]: [2001/02/21 16:26:40, 0] lib/util_sock.c:(47
4)

Feb 21 16:26:40 server3 smbd[1409]: read_socket_data: recv failure for 4. Erro
r = Connection reset by peer

[James A. Donovan]

Does this server have the same patches applied as the ones you're not having problems with?

You can get a list of the active patches by running:

# /usr/contrib/bin/show_patches

Especially, look for which ARPA/Transport patch is installed.

[Craig Gilmore]

It appears that you have 3 separate incidents. I'll try and address each:

The Jan 29th messages, Failed to set socket option, is due to a "bad file number." This does not mean that there is a problem, rather that the file number that was used for the call was not valid. Since socket options are being set, the file would be a socket file, or a network connection. It is probably a bad number because the socket connection has been reset and the file is no longer held open by the process. This would point to a connection problem.

The Feb 15th errors: Connection timeout. This means that the socket connection hit a timer and was dropped. What timer is not able to be determined from this message. This could be a configurable option, or it may also point to a connection or network problem.

The Feb 21st errors: Connection reset by peer.
This means that the device on the other side of the network connection sent a RESET, or "stop what you are doing and DIE." This is not a problem with the system that logs the message, rather it is the other side that has hung up in a semi-rude maner. This may be due to many various items. Too many retransmissions could cause a reset to be generated. A hard reboot of a PC by a user could cause a reset to be sent. But at this point I would not suspect the 9000.

If these 3 messages are just a small sample of the types of messages you are seeing, then there is not 1 point of failure that can be pointed to since you have 3 different issues. If any 1 of these areas has a great number of occurrances then you may have some concern. NONE of these messages directly point to a problem with Samba. Rather a couple of them point to issues with using the network connections on your system. I would check the network and it's connections first for any possible problem areas.