366

What can I do to configure SSH on both client and servers to prevent Write Failed: broken pipe errors? It often occurs if you sleep your client computer and resume later.

11
  • 8
    Nothing really. The session was interrupted, and the security of the session was compromised. If you don't put the comp to sleep you can set a keep alive time for the client to shoot a keep alive heart beat to the server, but if the system's going to sleep then there is nothing that can be done.
    – darkdragn
    Apr 28, 2012 at 23:44
  • 3
    In this case I'm looking for something that would allow me to re-initiate a broken ssh connection (based probably on the exit code) and restore using screen?
    – sorin
    Apr 30, 2012 at 7:11
  • 7
    You people are wrong: I have TWO desktop client machines connecting to the SAME server. One is ubuntu 12.10, Quantal, whose SSH client works well, and keeps the connection for hours. The other is Ubuntu 14.10, Utopic, just aside the other and in a fresh install; after a couple of minutes, it blocks itself with this message. The rest of the network functions in the machine are not interrupted. So no, it's neither a network problem, nor a server problem, but a specific SSH CLIENT software problem, which CAN be solved, opposite to what "darkdragan" dares to say, that "nothing can be done".
    – David L
    Dec 18, 2014 at 6:09
  • 3
    And indeed, as I said: people talk too much when they say "nothing can be done", just as @darkdragn dared. I read Aram Kocharyan 's answer, and I applied it: 20 minutes ago... I realized that in my old Quantal Ubuntu 12.10, I had applied that instruction in that file [I just checked], two years ago, and that was the reason of the stability there. I did it here, and in these last 20 minutes, the connection has been stable since then. So please, people: refrain yourselves when daring to think that "nothing can be done", and refrain even more when trying to leave that message to other people.
    – David L
    Dec 18, 2014 at 6:34
  • 12
    @DavidL you should read the questions better before ranting. Your problem is not the same as the OP's, who clearly mentions putting the computer to sleep. Which by the way only one of the answers address ("mosh"), and it was posted 2 years after the question. However the other answers do the next best thing, which is proposing solutions to cases that can be solved more easily, like yours. Chill out, don't be so stressed, ranting doesn't do any good around here...
    – msb
    Apr 8, 2015 at 21:40

10 Answers 10

345

I have tried this in /etc/ssh/ssh_config for Linux and Mac:

Host *
ServerAliveInterval 120

This is how often, in seconds, it should send a keepalive message to the server. If that doesn't work then train a monkey to press enter every two minutes while you work.

You could set either ServerAliveInterval in /etc/ssh/ssh_config of the client machine or ClientAliveInterval in /etc/ssh/sshd_config of the server machine. Try reducing the interval if you are still getting the error.

Configuration for a single user can be set in file ~/.ssh/config both on the server and client side. Make sure the file has correct permissions chmod 644 ~/.ssh/config.

19
  • 5
    I'm not on a Mac, but Ubuntu 12.04 and the file for this operating system also seems to be ~/.ssh/config.
    – H2ONaCl
    Dec 14, 2012 at 13:26
  • 7
    OS X 10.8.4 gives an error Bad configuration option: ClientAliveInterval
    – ohho
    Jul 15, 2013 at 4:46
  • 3
    I get that same Bad configuration option error on OSX 10.8.4. Jul 18, 2013 at 19:57
  • 12
    Generally, you put these two commands into different parts of the system. Only ServerAliveInterval on the OSX client side... and only ClientAliveInterval on the sshd config file...
    – ftrotter
    Oct 23, 2013 at 22:56
  • 6
    My monkey said to me: "Why don't you type yourself "top [ENTER]"
    – Augusto
    Jul 9, 2015 at 15:34
114

SSH sessions may break due to numerous and possibly unavoidable reasons.

A useful utility which can be used to mitigate problems caused by this is called screen. Screen is a powerful utility that allows you to control multiple terminals which will stay alive independently of the ssh session. For example, if you run screen in an ssh session you will see a new terminal open and you can use that to run jobs. Lets say your ssh session dies in the process. Running screen -d then screen -r will reopen the last session and you will be able to continue from there. Make sure you read some of the documentation before using it.

9
  • 7
    This is probably the best answer, I'm not sure why it's not voted up higher. The other "fixes" are helpful in the special case where you would actually care about maintaining an SSH connection, but in most use cases I imagine the real concern is that the intended processes continue to run, regardless of any client/server connection issues whatsoever. Jun 3, 2013 at 17:33
  • 23
    I would also add Tmux as an alternative to screen. I find it more versatile and stable than screen. Jan 5, 2015 at 15:28
  • 4
    just leaving this here for future reference – you can conveniently run screen -d -r to recover your last session.
    – doplumi
    Nov 21, 2016 at 7:41
  • 3
    Or simply screen -dr. Or screen -x depending on what you're planning on doing. The point is, one ought to know what all those switches do, so that one can use the appropriate ones and not just blindly follow internet people's suggestions. There's a nice compact summary available here: ss64.com/bash/screen.html
    – flith
    May 28, 2018 at 6:27
  • 3
    This is not an answer to the problem Apr 16, 2019 at 9:26
63

Client configuration

Try creating the file:

~/.ssh/config

Add the contents:

Host *
  ServerAliveInterval 30
  ServerAliveCountMax 5

Now ssh to your server and see if your problem is fixed. ClientAliveInterval option is only useful when configuring the ssh server (aka sshd), it does not change a thing on the ssh client side, so don't use it in the above configuration file.

This will send a hello-are-you-there signal to the server if no packets have been received in the preceding 30 seconds (as specified above). However, if the number of consecutive hello-are-you-there signals reach ServerAliveCountMax then ssh will disconnect from the server. This value is defaulting to 3 (so 3*30 = 90 seconds without server activity), increase it if it suits your needs. There are alot more config options to the .ssh/config file and you could read:

Using an SSH Config File

For more information on other options. You may not want to apply this to every server you connect to which this example will. Or restrain it to only a particular server by replacing the line Host * with Host <IP> (replace by an IP address, see ssh_config man page).

Server configuration

Similarly you can tell the server to be gentle with your clients. The configuration file is /etc/ssh/sshd_config.

ClientAliveInterval 20
ClientAliveCountMax 5

You can either deactivate it by setting ClientAliveInterval to 0 or tweak ClientAliveInterval and ClientAliveCountMax to set a maximum ssh client inactivity without responding to the probes. One advantage of this settings over TCPKeepAlive is that the signals are sent through the encrypted channels, so it is less likely to be spoofable.

9
  • It doesn't work. I am facing the same error again.
    – user997704
    Oct 6, 2013 at 4:59
  • 3
    Try it straight from the command line and go lower: ssh -o ServerAliveInterval=5 user@host
    – Matt
    Oct 6, 2013 at 10:46
  • Tried that too..doesn't work. I really don't know what is going on with my system
    – user997704
    Nov 15, 2013 at 8:39
  • 2
    It's ClientAliveCountMax, NOT ClientAliveMaxCount
    – David G
    Sep 5, 2015 at 1:11
  • @DavidG Please edit the answer with your corrections. Feb 3, 2016 at 17:35
34

I'm remotely upgrading an Ubuntu server from lucid to precise and lost the ssh connection in the middle of the upgrade with the message "Write failed. Broken pipe". ClientAliveInterval and ServerAliveInterval did nothing. The solution is to turn on TCPKeepAlive options in client ssh:

TCPKeepAlive yes

in

/etc/ssh/ssh_config

Hope this will help

28

For the client, edit your ~/.ssh/config (or /etc/ssh/ssh_config) file as follow:

Host *
  TCPKeepAlive yes
  ServerAliveInterval 120

TCPKeepAlive - Specifies whether the system should send TCP keepalive messages to the other side. If they are sent, death of the connection or crash of one of the machines will be properly noticed. However, this means that connections will die if the route is down temporarily, and some people find it annoying (The default is 'yes').

ServerAliveInterval - Sets a timeout interval in seconds after which if no data has been received from the server, ssh(1) will send a message through the encrypted channel to request a response from the server. The default is 0, indicating that these messages will not be sent to the server.


For the server, edit your /etc/ssh/sshd_config as:

ClientAliveInterval 600
ClientAliveCountMax 0

If you want ssh client to exit (timeout) automatically after 10 minutes (600 seconds).

ClientAliveCountMax – This indicates the total number of checkalive message sent by the ssh server without getting any response from the ssh client. Default is 3.

ClientAliveInterval – This indicates the timeout in seconds. After x number of seconds, ssh server will send a message to the client asking for response. Deafult is 0 (server will not send message to client to check.).


See also: What do the options ServerAliveInterval and ClientAliveInterval in sshd_config do, precisely?

1
  • Setting a ServerAliveCountMax higher than the default on the client should also help keep the connection live for slow connections. Oct 1, 2018 at 20:29
20

I absolutely love Mosh. I frequently ssh into a server, close my laptop and go to a cafe, open it up and carry on as if nothing changed.

Mosh (mobile shell)

Remote terminal application that allows roaming, supports intermittent connectivity, and provides intelligent local echo and line editing of user keystrokes.

Mosh is a replacement for SSH. It's more robust and responsive, especially over Wi-Fi, cellular, and long-distance links.

Mosh is free software, available for GNU/Linux, FreeBSD, Solaris, Mac OS X, and Android.

6

For me, I was getting Write failed: Broken pipe even when I was actively typing in vim or at the shell prompt. I couldn't browse the internet locally either for awhile either. (I was connecting remotely to Ubuntu using Terminal.)

Others in my network stream a lot of video from Netflix and other places. I can't prove it, but I suspect its an ISP or router issue. For example, Verizon and Netflix are pointing fingers at each other for their customer's network problems.

If you've got a dial-up connection and are streaming video or music with a simultaneous SSH or telnet connection, it's inevitable at some point you'll get a broken pipe message. Upgrading my ISPs broadband package seemed to make my broken connection less frequent.

6

I posted my answer here, as it was not a Ubuntu VM.

https://unix.stackexchange.com/questions/259225/packet-write-wait-broken-pipe-even-leaving-top-running

ssh -o IPQoS=throughput user@host
3

I have a script on the remote server that never seems to fails, regardless of the SSH configuration client or server.

#!/bin/bash
while true; do date; sleep 10; done;

Save it to some dummy.sh file and quickly run it before you minimize the window or move away from it. It will keep printing the current time stamp on the server and keeps your connection alive as long as the connection is not dropped by any other reason. When you get back to that terminal, just hit CTRL+C and keep working.

2
  • 12
    or simply leave top running
    – Eben Geer
    Dec 19, 2013 at 22:43
  • or run tail -f /dev/null
    – Mukundhan
    Mar 28, 2020 at 9:21
1

You can add these args every time you invoke ssh: -o ServerAliveInterval=15 -o ServerAliveCountMax=3

You don't have to edit /etc/ssh/*config files if you do this.

You can create an bash alias or function or script to make this easy.

E.g. these bash functions, you can add into your .bashrc, do_ssh is used manually to turn on keepalives. do_ssh_pty is used within scripts to set pty and avoid prompts.

do_ssh() {
    ssh -o ServerAliveInterval=15 -o ServerAliveCountMax=3 $*
}

do_ssh_pty() {
    ssh -tt -o "BatchMode=yes" -o "StrictHostKeyChecking=no" -o ServerAliveInterval=15 -o ServerAliveCountMax=3 $*
}

Now do_ssh user@host can be used or do_ssh user@host <args> <command> and keepalives will be active.

1
  • 2
    So for me the client side QoS setting didn't work but I compared sshd_config to a raspberry pi that did not have this and I found IPQoS 0x00 at the end of the config. I added this to my Ubuntu config, restarted sshd and now I'm good.
    – rjb101
    Apr 15, 2022 at 16:33

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .