Unfortunately it appears that my efforts to fix things didn't work... clients are still getting: [2001/01/23 03:37:46, 0] lib/util_sock.c:write_socket_data(540) write_socket_data: write failure. Error = Broken pipe [2001/01/23 03:37:46, 0] lib/util_sock.c:write_socket(566) write_socket: Error writing 4 bytes to socket 6: ERRNO = Broken pipe [2001/01/23 03:37:46, 0] lib/util_sock.c:send_smb(754) Error writing 4 bytes to client. -1. Exiting which results in the client being bounced. How can I make samba less paranoid about doing lookups? Currently it's using the default, which is host lmhost wins bcast . I ended up adding all of the ip's and computers to /etc/hosts , without luck. If i ping an IP address: [root@gene samba]# ping 192.168.0.56 Warning: no SO_TIMESTAMP support, falling back to SIOCGSTAMP PING 192.168.0.56 (192.168.0.56) from 192.168.0.1 : 56(84) bytes of data. 64 bytes from 192.168.0.56: icmp_seq=0 ttl=128 time=1.757 msec 64 bytes from 192.168.0.56: icmp_seq=1 ttl=128 time=481 usec whereas pinging on my own network, with some real names & addresses gives: [root@gemini /public]# ping 198.142.14.196 PING 198.142.14.196 (198.142.14.196) from 198.142.14.194 : 56(84) bytes of data. 64 bytes from celery.faredge.com.au (198.142.14.196): icmp_seq=0 ttl=128 time=0.5 ms 64 bytes from celery.faredge.com.au (198.142.14.196): icmp_seq=1 ttl=128 time=0.4 ms A couple of other comments - it doesn't appear to happen to NT clients - only to windows 9x clients. We usually use dhcp, and set: option domain-name "corbettresearch.com.au"; option domain-name-servers 192.168.0.1; option netbios-name-servers 192.168.0.1; option netbios-node-type 0x2; default-lease-time 604800; max-lease-time 604800; in dhcpd.conf. We map MAC addresses in the file below the global area and deny unknown hosts. All the computers are successfully being allocated IPs, and having the WINS server set etc. Other info: security = user no other NT servers only IP 100Mb hub hosts allow = 192.168.0. 127. encrypted passwords socket options = TCP_NODELAY SO_RCVBUF=8192 SO_SNDBUF=8192 interfaces = eth1 remote announce = 192.168.0 local master = yes os level = 33 domain master = yes preferred master = yes domain logons = yes # the default order is "host lmhosts wins bcast". "host" means use the unix ; name resolve order = host wins lmhosts bcast wins support = yes dns proxy = no all clients have service packs / security patches A couple of thoughts - if I'm using samba as a wins server, should I put wins first here? hosts contains an accurate list of machines, as does dhcpd.conf... I'm pretty close to trying anything here... Cheers, Chris
> [2001/01/23 03:37:46, 0] lib/util_sock.c:write_socket_data(540) > write_socket_data: write failure. Error = Broken pipe > [2001/01/23 03:37:46, 0] lib/util_sock.c:write_socket(566) > write_socket: Error writing 4 bytes to socket 6: ERRNO = Broken pipe > [2001/01/23 03:37:46, 0] lib/util_sock.c:send_smb(754) > Error writing 4 bytes to client. -1. ExitingI have the same error msgs: [2001/01/25 10:35:00, 0] lib/util_sock.c:(540) write_socket_data: write failure. Error = Broken pipe [2001/01/25 10:35:44, 0] lib/util_sock.c:(540) write_socket_data: write failure. Error = Broken pipe [2001/01/25 10:36:12, 0] lib/util_sock.c:(540) write_socket_data: write failure. Error = Broken pipe> How can I make samba less paranoid about doing lookups? Currently it's using > the default, which is host lmhost wins bcast . I ended up adding all of the > ip's and computers to /etc/hosts , without luck. If i ping an IP address:Just wondering, why do you think it is the lookups that causes these errors?> A couple of other comments - it doesn't appear to happen to NT clients - > only to windows 9x clients. > We usually use dhcp, and set:> I'm pretty close to trying anything here...I see that you have problems with only 9x clients. Isnt there an option in explorer for 9x clients that takes the network disks 'offline' after a certain amount of time? Would not this cause a "Broken pipe" whenever the disk goes offline after being idle for awhile? I have been ignoring these error-msgs (even though they are filling up my log.smb) because i am running a samba 2.0.7 binary-install and thought it might have todo with some preset TCP_KEEPALIVE compile option or something. If you figure it out I would like to hear what you did. Regards Per Kjetil Grotnes --- IT-Seksjonen, Plan- og bygningsetaten, Oslo Kommune Tlf: 22 66 26 61, Fax: 22 66 26 65
Chris Herrmann wrote:> > [2001/01/23 03:37:46, 0] lib/util_sock.c:write_socket_data(540) > write_socket_data: write failure. Error = Broken pipe....> How can I make samba less paranoid about doing lookups?...> hosts allow = 192.168.0. 127.Chris, I'm coming in late on this thread. However, if you want to prevent smbd from preforming gethostbyaddr() calls, remove the hosts allow line in your smb.conf. Otherwise, you should make sure that all clients have forward and reverse DNS entries. Cheers, jerry ---------------------------------------------------------------------- /\ Gerald (Jerry) Carter Professional Services \/ http://www.valinux.com/ VA Linux Systems gcarter@valinux.com http://www.samba.org/ SAMBA Team jerry@samba.org http://www.plainjoe.org/ jerry@plainjoe.org "...a hundred billion castaways looking for a home." - Sting "Message in a Bottle" ( 1979 )