Run sudo commands in Chef resource block - bash

To simplify, consider following block in a cookbook (cookbook-test) recipe.
79: bash 'Running sudo test sleep command' do
80: user 'root'
81: cwd '/tmp'
82: code <<-EOH
83: sudo sleep 1000
84: EOH
85: end
Running this as
"chef-client -o cookbook-test"
Output:
Mixlib::ShellOut::ShellCommandFailed
------------------------------------
Expected process to exit with [0], but received '1'
---- Begin output of "bash" "/tmp/chef-script20150813-3835-3kj758" ----
STDOUT:
STDERR: sudo: sorry, you must have a tty to run sudo
---- End output of "bash" "/tmp/chef-script20150813-3835-3kj758" ----
Ran "bash" "/tmp/chef-script20150813-3835-3kj758" returned 1
I added "sudo sleep" just to exemplify usecase. In real case, we run scripts inside above resource block, and these scripts has sudo commands.
After some debugging found that "bash" and "execute" resource blocks both do not have tty allocated to run commands inside them.
Please share your thoughts.

Here's the thing:
Any facility you could use in Chef to run sudo with an allocated tty could also be used by anybody else, which means the requiretty directive in your sudoers is effectively useless. So you might as well just remove it and save yourself the trouble of working around it.
Having said that, here are some ways to work around the problem:
Are you able to ssh to localhost without a password? You could just use ssh -tt localhost sudo somecomand ....
You can use the expect tool, which is designed for controlling terminal-oriented programs. Something like expect -c "spawn sudo somecommand; interact".
You can use screen, with something like screen sudo somecommand.

Related

Bash won’t recognize existence of closing single quote in systemd unit file

I have the following systemd unit file set to automatically update all Arch Linux and AUR packages at the same time (using the yay AUR helper, of course) while also attempting to temporarily add (and then delete after it’s done, for obvious reasons) a sudoers.d entry to briefly give nobody sudo access to pacman in order to get AUR packages updated:
[Unit]
Description=Automatic Update
After=network-online.target
[Service]
Type=simple
SyslogIdentifier=autoupdate
ExecStartPre=/bin/bash -c 'echo \'nobody ALL= NOPASSWD: /usr/bin/pacman\' > /etc/sudoers.d/autoupdate'
ExecStart=/bin/bash -c \”XDG_CACHE_HOME=/var/tmp PWD=/var/tmp sudo -E -u nobody yay -Syuq --noconfirm --devel --timeupdate\”
ExecStartPost=/usr/bin/rm -f /etc/sudoers.d/autoupdate
KillMode=process
KillSignal=SIGINT
[Install]
WantedBy=multi-user.target
The problem is that bash fails to acknowledge the existence of the closing single quote on the ExecStartPre line:
nobody: -c: line 1: unexpected EOF while looking for matching `’`
nobody: -c: line 2: syntax error: unexpected end of file
This is of course despite the fact that manually typing sudo bash -c ‘echo nobody ALL\=NOPASSWD: /usr/bin/pacman > /etc/sudoers.d/autoupdate’ into my shell succeeds without incident.
What could be causing this discrepancy?
Turns out the overcomplication of the issue was rooted in the use of ExecStartPost= instead of ExecStopPost=. Once I changed the former to the latter, the original version of the unit file from long before this was posted (which was far simpler) worked perfectly.
regardless of why you want to use sudo even though you are root..
and without thinking about your code..
Use a script instead
ExecStartPre=/path/to/your/script prestart
ExecStart=/path/to/your/script start
ExecStartPost=/path/to/your/script poststart
your script
#!/bin/bash
case $1 in
prestart) echo "nobody ALL= NOPASSWD: /usr/bin/pacman" > /etc/sudoers.d/autoupdate;;
start) XDG_CACHE_HOME=/var/tmp
PWD=/var/tmp
sudo -E -u nobody yay -Syuq --noconfirm --devel --timeupdate;;
poststart) rm -f /etc/sudoers.d/autoupdate;;
esac

What regex or command can I use to trim this output?

I am trying to capture a specific output in a $ variable for my remote server configuration that will run commands one after another.
In an ubuntu environment that has pm2 node package installed, it comes with a command that will output something I need to run.
Command 1:
PM2=$(pm2 startup systemd)
Will output this string when I run echo $PM2:
[PM2] Init System found: systemd [PM2] You have to run this command as root. Execute the following command: sudo env PATH=$PATH:/usr/bin /usr/lib/node_modules/pm2/bin/pm2 startup systemd -u username --hp /home/username
I need to capture this exact output as a $var:
sudo env PATH=$PATH:/usr/bin /usr/lib/node_modules/pm2/bin/pm2 startup systemd -u username --hp /home/username
So I can have my cloud init config file run it in the next command.
Command 2:
$PM2
How can I get $PM2 to only have the output value of
sudo env PATH=$PATH:/usr/bin /usr/lib/node_modules/pm2/bin/pm2 startup systemd -u username --hp /home/username
This may help :
pm2response=$(pm2 startup systemd) # Use lower case for user defined variables
${pm2response#*Execute the following command:} # Shell param expansion
But this assumes that your string has the phrase Execute the following command: in it though I guess I am right in assuming so. Good luck!
Note : More on SHELL Parameter Substitution [ here ]

Execute gcloud commands in a bash script

gcloud init command doesn't offer login prompt during a bash script execution.
But it offered the login after I typed exit command manually after script ended.
vagrant#vagrant-ubuntu-trusty-64:~$ exit
logout
Welcome! This command will take you through the configuration of gcloud.
Settings from your current configuration [default] are:
Your active configuration is: [default]
Pick configuration to use:
[1] Re-initialize this configuration [default] with new settings
[2] Create a new configuration
Please enter your numeric choice: 1
Your current configuration has been set to: [default]
To continue, you must log in. Would you like to log in (Y/n)?
My bash script:
#!/usr/bin/env bash
OS=`cat /proc/version`
function setupGCE() {
curl https://sdk.cloud.google.com | bash
`exec -l $SHELL`
`gcloud init --console-only`
`chown -R $USER:$USER ~/`
}
if [[ $OS == *"Ubuntu"* || $OS == *"Debian"* ]]
then
sudo apt-get -y install build-essential python-pip python-dev curl
sudo pip install apache-libcloud
setupGCE
fi
How can I get the login prompt during the bash script execution?
There are a number of issues with the posted snippet.
The correct snippet is (probably):
function setupGCE() {
curl https://sdk.cloud.google.com | bash
gcloud init --console-only
chown -R $USER:$USER ~/
}
The first error with the original, which you discovered yourself (the what of it at least it not the why), is that exec -l $SHELL is blocking progress. It does that because you've run an interactive shell that is now waiting on you for input and the function is waiting for that process to exit before continuing.
Additionally, exec replaces the current process with the spawned process. You got lucky here actually. Had you not wrapped the call to exec in single quotes your function would have exited the shell script entirely when you exited the $SHELL it launched. As it is, however, exec just replaced the sub-shell that the backticks added and so you were left with a child process that could safely exit and return you to the parent/main script.
The second issue is that backticks run the command they surround and then replace themselves with the output. This is why
echo "bar `echo foo` baz"
outputs bar foo baz, etc. (Run set -x before running that to see what commands are actually being run.) So when you write
`gcloud init --console-only`
what you are saying is "run gcloud init --console-only then take its output and replace the command with that" which will then attempt to run the output as a command itself (which is likely not what you wanted). Similarly on the other lines.
This happens to not have been problematic here though as chown and likely gcloud init don't return anything and so the resulting command line is empty.
Somehow the exec -l $SHELL did all the mess. I changed it to source ~/.bashrc and now it works.

TeamCity with Xcode3 and Xcode4 [duplicate]

I am trying to compile some sources using a makefile. In the makefile there is a bunch of commands that need to be ran as sudo.
When I compile the sources from a terminal all goes fine and the make is paused the first time a sudo command is ran waiting for password. Once I type in the password, make resumes and completes.
But I would like to be able to compile the sources in NetBeans. So, I started a project and showed netbeans where to find the sources, but when I compile the project it gives the error:
sudo: no tty present and no askpass program specified
The first time it hits a sudo command.
I have looked up the issue on the internet and all the solutions I found point to one thing: disabling the password for this user. Since the user in question here is root. I do not want to do that.
Is there any other solution?
Granting the user to use that command without prompting for password should resolve the problem. First open a shell console and type:
sudo visudo
Then edit that file to add to the very end:
username ALL = NOPASSWD: /fullpath/to/command, /fullpath/to/othercommand
eg
john ALL = NOPASSWD: /sbin/poweroff, /sbin/start, /sbin/stop
will allow user john to sudo poweroff, start and stop without being prompted for password.
Look at the bottom of the screen for the keystrokes you need to use in visudo - this is not vi by the way - and exit without saving at the first sign of any problem. Health warning: corrupting this file will have serious consequences, edit with care!
Try:
Use NOPASSWD line for all commands, I mean:
jenkins ALL=(ALL) NOPASSWD: ALL
Put the line after all other lines in the sudoers file.
That worked for me (Ubuntu 14.04).
Try:
ssh -t remotehost "sudo <cmd>"
This will remove the above errors.
After all alternatives, I found:
sudo -S <cmd>
The -S (stdin) option causes sudo to read the password from the standard input instead of the terminal device.
Source
Above command still needs password to be entered. To remove entering password manually, in cases like jenkins, this command works:
echo <password> | sudo -S <cmd>
sudo by default will read the password from the attached terminal. Your problem is that there is no terminal attached when it is run from the netbeans console. So you have to use an alternative way to enter the password: that is called the askpass program.
The askpass program is not a particular program, but any program that can ask for a password. For example in my system x11-ssh-askpass works fine.
In order to do that you have to specify what program to use, either with the environment variable SUDO_ASKPASS or in the sudo.conf file (see man sudo for details).
You can force sudo to use the askpass program by using the option -A. By default it will use it only if there is not an attached terminal.
Try this one:
echo '' | sudo -S my_command
For Ubuntu 16.04 users
There is a file you have to read with:
cat /etc/sudoers.d/README
Placing a file with mode 0440 in /etc/sudoers.d/myuser with following content:
myuser ALL=(ALL) NOPASSWD: ALL
Should fix the issue.
Do not forget to:
chmod 0440 /etc/sudoers.d/myuser
Login into your linux. Fire following commands. Be careful, as editing sudoer is a risky proposition.
$ sudo visudo
Once vi editor opens make the following changes:
Comment out Defaults requiretty
# Defaults requiretty
Go to the end of the file and add
jenkins ALL=(ALL) NOPASSWD: ALL
If by any chance you came here because you can't sudo inside the Ubuntu that comes with Windows10
Edit the /etc/hosts file from Windows (with Notepad), it'll be located at: %localappdata\lxss\rootfs\etc, add 127.0.0.1 WINDOWS8, this will get rid of the first error that it can't find the host.
To get rid of the no tty present error, always do sudo -S <command>
This worked for me:
echo "myuser ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
where your user is "myuser"
for a Docker image, that would just be:
RUN echo "myuser ALL=(ALL) NOPASSWD:ALL" >> /etc/sudoers
In Jenkins:
echo '<your-password>' | sudo -S command
Eg:-
echo '******' | sudo -S service nginx restart
You can use Mask Password Plugin to hide your password
Make sure the command you're sudoing is part of your PATH.
If you have a single (or multi, but not ALL) command sudoers entry, you'll get the sudo: no tty present and no askpass program specified when the command is not part of your path (and the full path is not specified).
You can fix it by either adding the command to your PATH or invoking it with an absolute path, i.e.
sudo /usr/sbin/ipset
Instead of
sudo ipset
Command sudo fails as it is trying to prompt on root password and there is no pseudo-tty allocated (as it's part of the script).
You need to either log-in as root to run this command or set-up the following rules in your /etc/sudoers
(or: sudo visudo):
# Members of the admin group may gain root privileges.
%admin ALL=(ALL) NOPASSWD:ALL
Then make sure that your user belongs to admin group (or wheel).
Ideally (safer) it would be to limit root privileges only to specific commands which can be specified as %admin ALL=(ALL) NOPASSWD:/path/to/program
I think I can help someone with my case.
First, I changed the user setting in /etc/sudoers referring to above answer. But It still didn't work.
myuser ALL=(ALL) NOPASSWD: ALL
%mygroup ALL=(ALL:ALL) ALL
In my case, myuser was in the mygroup.
And I didn't need groups. So, deleted that line.
(Shouldn't delete that line like me, just marking the comment.)
myuser ALL=(ALL) NOPASSWD: ALL
It works!
Running shell scripts that have contain sudo commands in them from jenkins might not run as expected. To fix this, follow along
Simple steps:
On ubuntu based systems, run " $ sudo visudo "
this will open /etc/sudoers file.
If your jenkins user is already in that file, then modify to look like this:
jenkins ALL=(ALL) NOPASSWD: ALL
save the file
Relaunch your jenkins job
you shouldnt see that error message again :)
This error may also arise when you are trying to run a terminal command (that requires root password) from some non-shell script, eg sudo ls (in backticks) from a Ruby program. In this case, you can use Expect utility (http://en.wikipedia.org/wiki/Expect) or its alternatives.
For example, in Ruby to execute sudo ls without getting sudo: no tty present and no askpass program specified, you can run this:
require 'ruby_expect'
exp = RubyExpect::Expect.spawn('sudo ls', :debug => true)
exp.procedure do
each do
expect "[sudo] password for _your_username_:" do
send _your_password_
end
end
end
[this uses one of the alternatives to Expect TCL extension: ruby_expect gem].
For the reference, in case someone else encounter the same issue, I was stuck during a good hour with this error which should not happen since I was using the NOPASSWD parameter.
What I did NOT know was that sudo may raise the exact same error message when there is no tty and the command the user try to launch is not part of the allowed command in the /etc/sudoers file.
Here a simplified example of my file content with my issue:
bguser ALL = NOPASSWD: \
command_a arg_a, \
command_b arg_b \
command_c arg_c
When bguser will try to launch "sudo command_b arg_b" without any tty (bguser being used for some daemon), then he will encounter the error "no tty present and no askpass program specified".
Why?
Because a comma is missing at the end of line in the /etc/sudoers file...
(I even wonder if this is an expected behavior and not a bug in sudo since the correct error message for such case shoud be "Sorry, user bguser is not allowed to execute etc.")
I was getting this error because I had limited my user to only a single executable 'systemctl' and had misconfigured the visudo file.
Here's what I had:
jenkins ALL=NOPASSWD: systemctl
However, you need to include the full path to the executable, even if it is on your path by default, for example:
jenkins ALL=NOPASSWD: /bin/systemctl
This allows my jenkins user to restart services but not have full root access
If you add this line to your /etc/sudoers (via visudo) it will fix this problem without having to disable entering your password and when an alias for sudo -S won't work (scripts calling sudo):
Defaults visiblepw
Of course read the manual yourself to understand it, but I think for my use case of running in an LXD container via lxc exec instance -- /bin/bash its pretty safe since it isn't printing the password over a network.
Using pipeline:
echo your_pswd | sudo -S your_cmd
Using here-document:
sudo -S cmd <<eof
pwd
eof
#remember to put the above two lines without "any" indentations.
Open a terminal to ask password (whichever works):
gnome-terminal -e "sudo cmd"
xterm -e "sudo cmd"
I faced this issue when working on an Ubuntu 20.04 server.
I was trying to run a sudo command from a remote machine to deploy an app to the server. However when I run the command I get the error:
sudo: no tty present and no askpass program specified
The remote script failed with exit code 1
Here's how I fixed it:
The issue is caused by executing a sudo command which tries to request for a password, but sudo does not have access to a tty to prompt the user for a passphrase. As it can’t find a tty, sudo falls back to an askpass method but can’t find an askpass command configured, so the sudo command fails.
To fix this you need to be able to run sudo for that specific user with no password requirements. The no password requirements is configured in the /etc/sudoers file. To configure it run either of the commands below:
sudo nano /etc/sudoers
OR
sudo visudo
Note: This opens the /etc/sudoers file using your default editor.
Next, Add the following line at the bottom of the file:
# Allow members to run all commands without a password
my_user ALL=(ALL) NOPASSWD:ALL
Note: Replace my_user with your actual user
If you want the user to run specific commands you can specify them
# Allow members to run specific commands without a password
my_user ALL=(ALL) NOPASSWD:/bin/myCommand
OR
# Allow members to run specific commands without a password
my_user ALL=(ALL) NOPASSWD: /bin/myCommand, /bin/myCommand, /bin/myCommand
Save the changes and exit the file.
For more help, read the resource in this link: sudo: no tty present and no askpass program specified
That's all.
I hope this helps
The solution to the problem is
If you came across this issue anywhere else apart from the Jenkins instance follow this from the 2nd step. The first step is for the user who is having issue with the Jenkins instance.
Go to Jenkins instance of Google Cloud Console.
Enter the commands
sudo su
visudo -f /etc/sudoers
Add following line at the end
jenkins ALL= NOPASSWD: ALL
Checkout here to understand the rootcause of this issue
No one told what could cause this error, in case of migration from one host to another, remember about checking hostname in sudoers file:
So this is my /etc/sudoers config
User_Alias POWERUSER = user_name
Cmnd_Alias SKILL = /root/bin/sudo_auth_wrapper.sh
POWERUSER hostname=(root:root) NOPASSWD: SKILL
if it doesn't match
uname -a
Linux other_hostname 3.10.17 #1 SMP Wed Oct 23 16:28:33 CDT 2013 x86_64 Intel(R) Core(TM) i3-4130T CPU # 2.90GHz GenuineIntel GNU/Linux
it will pop up this error:
no tty present and no askpass program specified
Other options, not based on NOPASSWD:
Start Netbeans with root privilege ((sudo netbeans) or similar) which will presumably fork the build process with root and thus sudo will automatically succeed.
Make the operations you need to do suexec -- make them owned by root, and set mode to 4755. (This will of course let any user on the machine run them.) That way, they don't need sudo at all.
Creating virtual hard disk files with bootsectors shouldn't need sudo at all. Files are just files, and bootsectors are just data. Even the virtual machine shouldn't necessarily need root, unless you do advanced device forwarding.
Although this question is old, it is still relevant for my more or less up-to-date system. After enabling debug mode of sudo (Debug sudo /var/log/sudo_debug all#info in /etc/sudo.conf) I was pointed to /dev: "/dev is world writable". So you might need to check the tty file permissions, especially those of the directory where the tty/pts node resides in.
I was able to get this done but please make sure to follow the steps properly.
This is for the anyone who is getting import errors.
Step1: Check if files and folders have got execute permission issue.
Linux user use:
chmod 777 filename
Step2: Check which user has the permission to execute it.
Step3: open terminal type this command.
sudo visudo
add this lines to the code below
www-data ALL=(ALL) NOPASSWD:ALL
nobody ALL=(ALL) NOPASSWD:/ALL
this is to grant permission to execute the script and allow it to use all the libraries. The user generally is 'nobody' or 'www-data'.
now edit your code as
echo shell_exec('sudo -u the_user_of_the_file python your_file_name.py 2>&1');
go to terminal to check if the process is running
type this there...
ps aux | grep python
this will output all the process running in python.
Add Ons:
use the below code to check the users in your system
cut -d: -f1 /etc/passwd
Thank You!
1 open /etc/sudoers
type sudo vi /etc/sudoers. This will open your file in edit mode.
2 Add/Modify linux user
Look for the entry for Linux user. Modify as below if found or add a new line.
<USERNAME> ALL=(ALL) NOPASSWD: ALL
3 Save and Exit from edit mode
I had the same error message when I was trying to mount sshfs which required sudo : the command is something like this :
sshfs -o sftp_server="/usr/bin/sudo /usr/lib/openssh/sftp-server" user#my.server.tld:/var/www /mnt/sshfs/www
by adding the option -o debug
sshfs -o debug -o sftp_server="/usr/bin/sudo /usr/lib/openssh/sftp-server" user#my.server.tld:/var/www /mnt/sshfs/www
I had the same message of this question :
sudo: no tty present and no askpass program specified
So by reading others answer I became to make a file in /etc/sudoer.d/user on my.server.tld with :
user ALL=NOPASSWD: /usr/lib/openssh/sftp-server
and now I able to mount the drive without giving too much extra right to my user.
Below actions work for on ubuntu20
edit /etc/sudoers
visudo
or
vi /etc/sudoers
add below content
userName ALL=(ALL) NOPASSWD: ALL
%sudo ALL=(ALL:ALL) NOPASSWD:ALL
I'm not sure if this is a more recent change, but I just had this problem and sudo -S worked for me.

Running individual commands works but not when combines in a shell script.. Why?

I need to run the following set of commands in a shell script
modprobe nbd
sudo qemu-nbd -c /dev/nbd0 path/to/image/file
sudo mount /dev/nbd0p1 /mnt/temp
python copyFiles.py
sudo umount /mnt/temp
sudo qemu-nbd -d /dev/nbd0
sudo rmmod nbd
When I individually run these commands it works fine, but when I put them in a shell script and executed that shell script, I always end up with an error in the mount command.
So I threw in a sleep 1 before mount and it works as expected.
What could be the reason behind this?
(Some sort of asynchronous call registration delay/ race condition?)
mount error: mount point /mnt/temp does not exist
So it seems the directory /mnt/temp doesn't exist when you are running it as a shell script. Just create it or add this in your script somewhere before the mount command:
mkdir /mnt/temp 2>&1 /dev/null
Both mount and the previous command require escalated privileges. Does it error cause the lock is still in place from the previous command when mount tries to run?

Resources