RE: lost tcp/ip connections
Donnie, You are probably aware of this, but still... To avoid loosing sessions due to network problems, I screen them. I added exec screen -xRR -T vt100 to my .profile script. Of course you should change the options to suit your needs. Now whenever I get disconnected, and then reconnect - I do not loose anything. It works for me, and helped me run some lengthy reports remotely. BTW, I use ssh instead of telnet. JD On Fri, 30 Jan 2004, Donnie Jacobs wrote: Hi Robert, Thanks for the information...I've tried what you suggested. I modified the tcp_keepalive_time, tcp_keepalive_intvl, and tcp_keepalive_probe settings. Below is how they're currently set [EMAIL PROTECTED] /root]# more /proc/sys/net/ipv4/tcp_keepalive_* :: /proc/sys/net/ipv4/tcp_keepalive_intvl :: 30 :: /proc/sys/net/ipv4/tcp_keepalive_probes :: 5 :: /proc/sys/net/ipv4/tcp_keepalive_time :: 60 I set these up in the init script and rebooted the machine, verified that they were still in place and then did the whole connect / unplug cable / reconnect cable bit. Based on these settings, I would have thought that after 60 seconds of inactivity, that I would get a max of 5 probes at 30 second intervals, and if no response was received, the connection would die. That should be a total of...60 seconds + (30 seconds * 5 tries = 150) = 210 seconds or just under 4 minutes... However, I've been back in for about 15 minutes now, and still see both logins when I do the who Did I miss something? Thanks, Donnie Jacobs Sr Developer GC Services LP 713-776-6503 [EMAIL PROTECTED] -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] Behalf Of Robert Porter Sent: Friday, January 30, 2004 1:19 PM To: [EMAIL PROTECTED] Subject: Re: lost tcp/ip connections That's a TCP/IP keepalive timeout... The setting is viewable and changable - It should be found at /proc/sys/net/ipv4/tcp_keepalive_time Changing it won't survive a reboot though. You'll either need to change it every time or do so in a script at init. rfp [EMAIL PROTECTED] 01/30/04 01:12PM Hello all, I'm wondering if there's a way within Linux to detect when a tcp/ip connection is lost. Here's the scenario... I telnet to my linux machine from my PC. I then disconnect my Ethernet cable (simulating a network failure, loss of broadband service, etc..) Then I reconnect my Ethernet cable and telnet into the linux machine again. If I do a who command, I see the old connection as well as my current connection. Running RedHat Linux 8, universe 10.0.13 I've tried several variations of netstat, who, etc...to try and identify these stranded connections, with no luck. Any ideas? Thanks, Donnie Jacobs Sr Developer GC Services LP 713-776-6503 [EMAIL PROTECTED] ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users
RE: lost tcp/ip connections
It's available for aix. Please check: http://www-1.ibm.com/servers/aix/products/aixos/linux/rpmgroups.html JD On Mon, 2 Feb 2004 [EMAIL PROTECTED] wrote: What is 'screen'? I don't find it on AIX. Is it a Linux thing? If it does what I think you are implying, it would be a wonderful thing. Thanks Harman Armstrong [EMAIL PROTECTED] JD [EMAIL PROTECTED] Sent by: [EMAIL PROTECTED] 02/02/2004 01:22 PM Please respond to U2 Users Discussion List To: U2 Users Discussion List [EMAIL PROTECTED] cc: Subject:RE: lost tcp/ip connections Donnie, You are probably aware of this, but still... To avoid loosing sessions due to network problems, I screen them. I added exec screen -xRR -T vt100 to my .profile script snip ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users
RE: lost tcp/ip connections
There is a -n option on the telnetd command that says disable tcp keep-alives The default is for the keep-alives to be enabled. If I look at the /etc/xinetd.d/telnet file, I see the following : [EMAIL PROTECTED] /root]# cat /etc/xinetd.d/telnet # default: on # description: The telnet server serves telnet sessions; it uses \ # unencrypted username/password pairs for authentication. service telnet { flags = REUSE socket_type = stream wait = no user = root server = /usr/sbin/in.telnetd log_on_failure += USERID disable = no } based on the value of the server entry, it doesnt appear that the -n option should be coming into play. Donnie Jacobs Sr Developer GC Services LP 713-776-6503 [EMAIL PROTECTED] -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Robert Porter Sent: Friday, January 30, 2004 3:09 PM To: [EMAIL PROTECTED] Subject: RE: lost tcp/ip connections Are you logging in by telnet or ssh? I'm wondering if your daemon is doing something funny with keepalive for some reason. Some telnetd's do have an option to turn it off. Try a man in.telnetd or just man telnetd to see if that's a possibility. rfp [EMAIL PROTECTED] 01/30/04 02:37PM Hi Robert, Thanks for the information Ive tried what you suggested. I modified the tcp_keepalive_time, tcp_keepalive_intvl, and tcp_keepalive_probe settings. Below is how theyre currently set . [EMAIL PROTECTED] /root]# more /proc/sys/net/ipv4/tcp_keepalive_* :: /proc/sys/net/ipv4/tcp_keepalive_intvl :: 30 :: /proc/sys/net/ipv4/tcp_keepalive_probes :: 5 :: /proc/sys/net/ipv4/tcp_keepalive_time :: 60 I set these up in the init script and rebooted the machine, verified that they were still in place and then did the whole connect / unplug cable / reconnect cable bit. Based on these settings, I would have thought that after 60 seconds of inactivity, that I would get a max of 5 probes at 30 second intervals, and if no response was received, the connection would die. That should be a total of 60 seconds + (30 seconds * 5 tries = 150) = 210 seconds or just under 4 minutes However, Ive been back in for about 15 minutes now, and still see both logins when I do the who . Did I miss something? Thanks, Donnie Jacobs Sr Developer GC Services LP 713-776-6503 [EMAIL PROTECTED] -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]On Behalf Of Robert Porter Sent: Friday, January 30, 2004 1:19 PM To: [EMAIL PROTECTED] Subject: Re: lost tcp/ip connections That's a TCP/IP keepalive timeout... The setting is viewable and changable - It should be found at /proc/sys/net/ipv4/tcp_keepalive_time Changing it won't survive a reboot though. You'll either need to change it every time or do so in a script at init. rfp [EMAIL PROTECTED] 01/30/04 01:12PM Hello all, I'm wondering if there's a way within Linux to detect when a tcp/ip connection is lost. Here's the scenario... I telnet to my linux machine from my PC. I then disconnect my Ethernet cable (simulating a network failure, loss of broadband service, etc..) Then I reconnect my Ethernet cable and telnet into the linux machine again. If I do a who command, I see the old connection as well as my current connection. Running RedHat Linux 8, universe 10.0.13 I've tried several variations of netstat, who, etc...to try and identify these stranded connections, with no luck. Any ideas? Thanks, Donnie Jacobs Sr Developer GC Services LP 713-776-6503 [EMAIL PROTECTED] ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users ___ u2-users mailing list [EMAIL PROTECTED] http://www.oliver.com/mailman/listinfo/u2-users