I use vscode with remote-ssh to connect my server, after configuring, I want to connect my host, but it failed, the dialog box display:"could not establish connection to XX, The process tried to write to a nonexistent pipe."
output:
[16:45:20.916] Log Level: 3
[16:45:20.936] remote-ssh#0.49.0
[16:45:20.936] win32 x64
[16:45:20.944] SSH Resolver called for "ssh-remote+aliyun", attempt 1
[16:45:20.945] SSH Resolver called for host: aliyun
[16:45:20.945] Setting up SSH remote "aliyun"
[16:45:21.012] Using commit id "c47d83b293181d9be64f27ff093689e8e7aed054" and quality "stable" for server
[16:45:21.014] Install and start server if needed
[16:45:21.019] Checking ssh with "ssh -V"
[16:45:21.144] > OpenSSH_for_Windows_7.7p1, LibreSSL 2.6.5
[16:45:21.214] Running script with connection command: ssh -T -D 5023 aliyun bash
[16:45:21.221] Terminal shell path: C:\WINDOWS\System32\cmd.exe
[16:45:21.504] >
>
>
> ]0;C:\WINDOWS\System32\cmd.exe
[16:45:21.505] Got some output, clearing connection timeout
[16:45:21.577] >
>
>
>
[16:45:21.592] > Bad owner or permissions on C:\\Users\\DY/.ssh/config
>
[16:45:21.689] > The process tried to write to a nonexistent pipe.
>
[16:45:22.091] "install" terminal command done
[16:45:22.092] Install terminal quit with output: The process tried to write to a nonexistent pipe.
[16:45:22.093] Received install output: The process tried to write to a nonexistent pipe.
[16:45:22.096] Resolver error: The process tried to write to a nonexistent pipe
[16:45:22.107] ------
Add the absolute file path to a custom SSH config file(C:\Users\{USERNAME}\.ssh\config), and my problem is solved.
If you format/re-install Server OS, but use same IP as before,
you may encounter fingerprint mismatched.
You may need to delete old fingerprint in this file:
C:\Users\xxx.ssh\known_host
and old IP in the file:
C:\Users\xxx.ssh\config
Then try to add host again.
What worked for me:
delete ssh config folder both in C:\Program Data\ssh and C:\<user>\.ssh
In VS Code, press F1, choose Remote-SSH: Connect to Host...
Do NOT enter anything in the prompt, but instead choose + Add New SSH Host..
Enter the full ssh command, including the key (in case of Windows,
you may want to enclose the path with double quote mark) ssh -i "C:\path\to\key" user#host. (you need to make sure the key has a limited permission. Remove all inherited permissions, and only give a full control to the owner.)
You will be asked to choose a folder in which a new config file will be created. Choose any of the two options.
There will a prompt notifying that the new config file
has been created. Click connect
At least three things may be happening:
Option 1
The location of your config file is not the absolute location, meaning you are probably using the location of the folder where the config file is.
If that is the case, access your User Settings in VSCode. Scroll to the Extensions>Remote - SSH. And add config at the end of the absolute file path of your custom SSH config file. In Windows, it can be
C:\Users\user\.ssh\config
See image below
Option 2
Authentication problems.
If that is the case, one of the things that may solve is generating new SSH keys.
In Windows, for that, I recommend using MobaXterm.
In MobaXterm, open a new terminal and write
ssh-keygen -b 4096 -t rsa
Then, in the config file, make sure that the IdentityFile points to the location of the key. MobaXterm's home directory, usually, is C:\Users\user\Documents\MobaXterm. If it makes it easy, one can copy/move the keys to C:\Users\user\.ssh and then just add, in the config file, IdentityFile ~/.ssh/KEY_rsa (where KEY_rsa is the name of the [public] key).
Note that if you use PuTTY to generate the keys, on the server OpenSSH authorized_keys file, one doesn't want the public key that one saves, but the one that appears on top (see image bellow):
Option 3
Your config file may be wrong.
The config file tends to look as follows. Double check if the fields have the information needed for the connection to be established.
Host Test # This is the name we want to give the host
User user # This is the username
Hostname blabla.com # This is the hostname
PreferredAuthentications publickey
IdentityFile ~/.ssh/KEY_rsa # This is the location of the key
IdentitiesOnly yes
Port 50 # This varies
What worked for me was to delete all of the contents of folder: C:\Users\MYNAME.ssh That meant to delete both the config file and known-hosts. The config was probably the most important one to delete.
The solution in my case was editing the json settings file for VSC as shown here: https://code.visualstudio.com/docs/remote/troubleshooting#_troubleshooting-hanging-or-failing-connections
In VSC go to File, Preferences, Setting and click on the upper right hand icon (Open Settings, JSON). Add these two lines to settings.json and retry connecting:
"remote.SSH.showLoginTerminal": true,
"remote.SSH.useLocalServer": false
In my case I had another setup:
Git bash in Windows was configured and I am using the ssh.exe provided by this tool
In the "Remote SSH" extension in VSCode, I specified the full path of this ssh.exe
I am using multiple servers (with ProxyJump)
The error message is the same as the OP but in the logs it was written that the ssh config file was not found, where all the folder names was concatenated (because it did not recognize the windows path separator)
Problem: the ssh.exe is using a different path convention thant VSCode. ssh.exe is using the "/c/Users/..." pattern and VSCode is using the "C:\Users..." pattern.
Solution:
Make sure the SSH config is at a standard place (C:\Users\LOGIN\.ssh\config)
Remove the absolute path of the config file in the "Remote SSH" settings in VSCode
VSCode will still be able to access the settings using the standard path, and the ssh.exe configuration will still look at the same standard path so the connexion is working.
Note:
I have the error only when connecting with multiple ssh servers (using ProxyJump). When connecting only to the first server, the solution of #pszrux and this one are both working for me.
This is probably something everyone has tried before looking here, but it worked for me. The server I was trying to ssh into was not responding, leading to the nonexistant pipe error. I rebooted the server and everything worked fine.
OS: windows 10
In my case there were permission issues. Repeatedly changing inheritance in windows did not solve the issue. Finally this worked
change the folder in which the config file is stored.
From C:/users/usr/.ssh/config to D:/config
and changed the config path in vscode remote ssh settings.
This worked for me.
This seems to be a problem with varied causes and corresponding remedies. In my case the problem had to do with the version of ssh I was using. In my Windows path there were two places were an instance of ssh.exe resided:
C:\Program Files (x86)\OpenSSH\bin
C:\WINDOWS\System32\OpenSSH\
After using both paths to set the "Remote.SSH: Path" parameter (which is in "Remote.SSH: Settings" [see here]), i.e. first C:\Program Files (x86)\OpenSSH\bin\ssh.exe and then C:\WINDOWS\System32\OpenSSH\ssh.exe, the problem still persisted.
Then I looked at this and tried the git-provided ssh.exe, which I already had on my system (otherwise, just install git, it's good stuff anyway :) )
Setting the SSH path parameter with that version did the trick for me, i.e. setting path to:
C:\Program Files\Git\usr\bin\ssh.exe
In my case, I did what dalilander said, but instead of deleting the entire '.ssh' folder, I just needed to delete the file 'known_hosts' and then it worked. So the servers I had saved were not deleted.
The path of that folder is C:\Users\yourUsername\.ssh
For Windows:
Adding the escape character before the private key file name & using quotes around the path solved my issue.
//config file
Host 12.12.12.12
HostName 12.12.12.12
IdentityFile "C:\Users\USERNAME/\PRIVATEKEY" <----Note /\
User username
Trying to add the full path in "IdentityFile" made the trick
" IdentityFile C:\Users\xxx.ssh\xxx"
The solution below may be the last resort but it perfectly solved the issue for me in a Windows 10 local machine. I simply delete the known_hosts file under the directory C:\Users\[your-username]\.ssh, relaunch VS Code and reconnect the remote server through Remote Explorer. Everything works normally afterward.
This seems to be a general error when the ssh connection fails for one of a multitude of reasons.
Adding what my issue was, and what helped, because I don't see it in the other answers in here: I had re-installed the box I was connecting to, and with it, reset the key it was using to authenticate. The ssh process tried to connect and failed with the usual "someone might be MITMing you this very moment, the identification changed" error, visible in the VSCode terminal. Solution was to go to my authorized_keys file and remove the offending key.
Obviously only know that if you know for sure why the identification changed, and that it's harmless. Don't actually get MITMed.
I had this problem once.
All you need to do is,
Go to /Users/XXX/.ssh
if you are on the windows, use command : "del /f known_hosts" to delete the known_hosts on the command prompt.
3.Then go to C:\Users\XXX.ssh\config on the vs code( config file )
4.Delete the host and the user if the host that you are trying to connect to is already there.
5.Then try to connect to the new host as usual.This will work.
The problem here could the mismatch of the finger prints once you reinstall the OS o n your host machine.
So to solve this problem by deleting the host that was saved.
once the config file on the vs code is edited it should look like..below picture is to show how the config file should look(after deleting the host saved)
If you're using WSL and might think that you should update ~/.ssh/config, that might not be the case.
Copy the content from ~/.ssh/config
Append it to C:\User\xxx\.ssh\config windows file
Make sure the public/private key is on C:\User\xxx\.ssh\ and is listed in config
Reconnect
Had an existing(working) configuration and had the same error when I added a new one. What worked for me is instead of just adding a new host configuration, I also commented out the first working config. Didn't know what happened but it worked.
In my case this was an offending key in my known_hosts in Windows (vscode on windows, remote developing via ssh on linux).
The error that comes back in vscode is not explaining in any way.
In my case, the path to key file was wrong.
For me, (windows) the permissions on the .pem file were the Critical issue. I had Administrator group only on the pem file and it was not working. I had to explicitly add the Admin user as well (even though admin is of course in administrator group).
In my case, I had no internet connection.
I was connecting to the server via VPN but the remote configuration was incorrect and I couldn't access the server. (DNS related issues) The connection indicator was showing no errors, so I didn't think of that at first.
Oops :)
I really didn't want to delete my C:Users\valo\.ssh\config, so I played a little with the various entries. It turned out that for me the option IdentitiesOnly yes was the problem. I also disabled security inheritance on all key files in the .ssh folder and left only myself, with Full Rights. Here is what my C:Users\valo\.ssh\config looks like now:
CanonicalizeHostname yes
Host aws.r3
HostName 3.31.45.216
ForwardAgent yes
User ubuntu
IdentityFile C:Users\valo\.ssh\u1-client-20210203-090555.pem
# IdentitiesOnly yes # VSCode Remote doesn't like this flag...?
Host github
HostName github.com
User git
IdentityFile ~/.ssh/id_val_ed25519
Host github.vm
HostName github.com
User git
IdentityFile ~/.ssh/id_valo_ed25519
Host *
ForwardAgent yes
AddKeysToAgent yes
LogLevel FATAL
ChallengeResponseAuthentication no
Now I can connect to aws.r3 with VSCode Remote.
A possible solution:
First run cat $HOME/.ssh/id_rsa.pub on your computer. That will get you a key. Save this key somewhere.
Then open this file by running vim $HOME/.ssh/authorized_keys on the computer that you're are ssh'ing to. Then copy the key in a new line of this file and close it by typing :wq.
You are all set.
In my case it was because the name I gave the host in config was myuser#myhostname. So if your config file looks like this:
Host myuser#myhostname
HostName 12.64.88.234
User myuser
Port 22
Try changing it to this:
Host myuser
HostName 12.64.88.234
User myuser
Port 22
Mine was solved by adding ".pem" extension while specifying the private key.
Here's the sample config text for reference.
Host ec2-3-234-8-176.compute-1.amazonaws.com
HostName ec2-3-234-8-176.compute-1.amazonaws.com
IdentityFile ~/.ssh/privatekey.pem
User username
There can be several reasons that have nothing to do with the accepted answer. For me:
Ubuntu 14.04.1 LTS didn't seem to work
Issues with EC2 auth, see pem file config and pem file permissions
And for yet another seeming cause/solution:
Adding the config path explicitly to settings only caused an EISDIR error.
Removing the listing from known_hosts made it need to confirm the fingerprint, but it didn't provide a way to do that. I could see it trying in the terminal output.
The solution a coworker recommended was to delete the vscode-server files from the server. That was the key step in my case. But...
Connecting to the server using another client, I attempted to rm -rf ~/.vscode-server. I could not delete many of the files because "device or resource busy".
That eventually required doing the following:
fuser ~/.vscode-server/[one of the child files ...]. But, you can probably skip this, because mainly I needed to know what to search for. Plus, fuser and lsof were finicky about returning results -- they often just sit waiting for something, I don't know what.
ps -e | grep node since node is the running process using vscode-server files.
For each PID in the list of node processes from step 2, run ps -o user= -p PID, substituting PID with each process PID in turn. This creates a formatted list of the process's associated user.
This is to determine which node processes you own, so you're not even trying to kill anybody else's.
Starting with the lowest node PID I own, run kill -9 PID. You won't need to run this for all PIDs, because killing a lower PID killed a child PID after a few seconds. So keep checking which node processes still exist after killing one: ps -e | grep node
Finally, once all mine are killed, I can rm -rf ~/.vscode-server
Then, I was able to connect via ssh in VS Code again. And, since I left the fingerprint removed from known_hosts, it even asked to confirm the connection to the server (up top, in the command prompt).
Also, for reference, I left the remote-ssh: settings config file entry, mentioned in other solutions, blank.
For reference or further explanation of certain steps above (I don't intend to elaborate more than I did):
rm: cannot remove ‘.vscode-server/bin/xxxxxx/.nfs000000000xxxxxxxxxxxx’: Device or resource busy
How find out which process is using a file in Linux?
https://unix.stackexchange.com/questions/284934/return-owner-of-process-given-pid/284938
In my case it worked when I added the port in my expression, eg ssh user#host-or-ip -p 22. With '22' the default port number, but you can check which port the ssh system is listening on with the sudo service ssh status command.
WSL Related
In my case, the issue was that my keys were set up on Ubuntu on WSL and VS Code was looking for them in Windows. I copied the keys over from the WSL side to Windows and voila! Worked like a charm!
Steps that I took.
Navigated to the /home//.ssh folder in WSL and then entered explorer.exe .
From there, I copied the id_rsa and id_rsa.pub files and pasted them in the windows side, under C:\Users<username>.ssh
Then I tried connecting again from VS Code and it worked perfectly.
This is a known issue and I found a solution but it's not working for me.
First I had:
fatal: [openshift-node-compute-e50xx] => SSH Error: ControlPath too long
It is sometimes useful to re-run the command using -vvvv, which prints SSH debug output to help diagnose the issue.
So I created a ~/.ansible.cfg. The content of it:
[ssh_connection]
control_path=%(directory)s/%%h‐%%r
But after rerunning my ansible I stil have an error about 'too long'.
fatal: [openshift-master-32axx] => SSH Error: unix_listener: "/Users/myuser/.ansible/cp/ec2-xx-xx-xx-xx.eu-central-1.compute.amazonaws.com-centos.AAZFTHkT5xXXXXXX" too long for Unix domain socket
while connecting to 52.xx.xx.xx:22
It is sometimes useful to re-run the command using -vvvv, which prints SSH debug output to help diagnose the issue.
Why is it still too long?
The limit is 104 or 108 characters. (I found different statements on the web)
You XXXed out some sensitive information in the error message so it's not clear how long your path actually is.
I guess %(directory)s is replaced with the .ansible directory in your users folder. Removing that and using directly your user folder would save you 12 characters:
control_path=~/%%h‐%%r
Sure, that will spam your home directory with control sockets.
Depending on the actual length of your username, you could see if you can just create another directory or find a shorter path anywhere. For example, I use ~/.ssh/tmp/%%h_%%r
Only 3 chars less but it's enough.
Finally if none of that helps, you still could fall back using /tmp for storing the sockets. But be aware that anyone with access to /tmp on that machine might be able to use your sockets then.
Customizing the control_path solves the problem for me. Here is how to do it without spamming the home directory.
The control_path defaults to (documentation):
control_path=%(directory)s/ansible-ssh-%%h-%%p-%%r
Edit ansible config.
vim ~/.ansible.cfg
Here are sample file contents with new control_path value:
[defaults]
inventory=/etc/ansible/hosts
[ssh_connection]
control_path=%(directory)s/%%h-%%r
control_path_dir=~/.ansible/cp
Just to add more, as the error shows this problem generally happens when the control path is too long for Unix domain socket, hence, to specific to ansible.
You can easily fix this by updating your config file to use the %C format instead of %r#%h:%p as follow:
$ mkdir ~/.ssh/control
$ vim ~/.ssh/config
Host *
ControlPath ~/.ssh/control/%C
ControlMaster auto
More Detail: man ssh_config defines the %C format as 'a hash of the concatenation: %l%h%p%r'. And refer here.
For me, Ansible config file was missing. After that it worked for me.
I'm using fabric to remotely start a micro aws server, install git and a git repository, adjust apache config and then restart the server.
If at any point, from the fabfile I issue either
sudo('service apache2 restart') or run('sudo service apache2 restart') or a stop and then a start, the command apparently runs, I get the response indicating apache has started, for example
[ec2-184-73-1-113.compute-1.amazonaws.com] sudo: service apache2 start
[ec2-184-73-1-113.compute-1.amazonaws.com] out: * Starting web server apache2
[ec2-184-73-1-113.compute-1.amazonaws.com] out: ...done.
[ec2-184-73-1-113.compute-1.amazonaws.com] out:
However, if I try to connect, the connection is refused and if I ssh into the server and run
sudo service apache2 status it says that "Apache is NOT running"
Whilst sshed in, if run
sudo service apache start, the server is started and I can connect. Has anyone else experienced this? Or does anyone have any tips as to where I could look, in log files etc to work out what has happened. There is nothing in apache2/error.log, syslog or auth.log.
It's not that big a deal, I can work round it. I just don't like such silent failures.
Which version of fabric are you running?
Have you tried to change the pty argument (try to change shell too, but it should not influence things)?
http://docs.fabfile.org/en/1.0.1/api/core/operations.html#fabric.operations.run
You can set the pty argument like this:
sudo('service apache2 restart', pty=False)
Try this:
sudo('service apache2 restart',pty=False)
This worked for me after running into the same problem. I'm not sure why this happens.
This is an instance of this issue and there is an entry in the FAQ that has the pty answer. Unfortunately on CentOS 6 doesn't support pty-less sudo commands and I didn't like the nohup solution since it killed output.
The final entry in the issue mentions using sudo('set -m; service servicename start'). This turns on Job Control and therefore background processes are put in their own process group. As a result they are not terminated when the command ends.
When connecting to your remotes on behalf of a user granted enough privileges (such as root), you can manage system services as shown below:
from fabtools import service
service.restart('apache2')
https://fabtools.readthedocs.org/en/0.13.0/api/service.html
P.S. Its requires the installation of fabtools
pip install fabtools
Couple of more ways to fix the problem.
You could run the fab target with --no-pty option
fab --no-pty <task>
Inside fabfile, set the global environment variable always_use_pty to False, before your target code executes
env.always_use_pty = False
using pty=False still didn't solve it for me. The solution that ended up working for me is doing a double-nohup, like so:
run.sh
#! /usr/bin/env bash
nohup java -jar myapp.jar 2>&1 &
fabfile.py
...
sudo("nohup ./run.sh &> nohup.out", user=env.user, warn_only=True)
...