Configuring a new Digital Ocean droplet with SSH keys. When I run ssh-copy-id this is what I get:

ssh-copy-id user@012.345.67.89
/usr/bin/ssh-copy-id: INFO: attempting to log in with the new key(s), to filter out any that are already installed
/usr/bin/ssh-copy-id: INFO: 1 key(s) remain to be installed -- if you are prompted now it is to install the new keys
sign_and_send_pubkey: signing failed: agent refused operation
user@012.345.67.89's password: 

Number of key(s) added: 1

Now try logging into the machine, with:   "ssh 'user@012.345.67.89'"
and check to make sure that only the key(s) you wanted were added.

However, when I then attempt to ssh in, this happens:

ssh user@012.345.67.89
sign_and_send_pubkey: signing failed: agent refused operation
user@012.345.67.89's password: 

Upon entering the password, I am logged in just fine, but this of course defeats the purpose of creating the SSH key in the first place. I decided to take a look at the ssh-agent server-side and here's what I get:

user@012.345.67.89:~# eval `ssh-agent -s`
Agent pid 5715
user@012.345.67.89:~# ssh-add -l
The agent has no identities.

user/.ssh/authorized_keys does contain an ssh-rsa key entry, as well, but find -name "keynamehere" returns nothing.

Ответы (18)

Run ssh-add on the client machine, that will add the SSH key to the agent.

Confirm with ssh-add -l (again on the client) that it was indeed added.

I got a sign_and_send_pubkey: signing failed: agent refused operation error as well. But in my case the problem was a wrong pinentry path.

In my ${HOME}/.gnupg/gpg-agent.conf the pinentry-program property was pointing to an old pinentry path. Correcting the path there and restarting the gpg-agent fixed it for me.

I discovered it by following the logs with journalctl -f. There where log lines like the following containing the wrong path:

Jul 02 08:37:50 my-host gpg-agent[12677]: ssh sign request failed: No pinentry 
Jul 02 08:37:57 my-host gpg-agent[12677]: can't connect to the PIN entry module '/usr/local/bin/pinentry': IPC connect call failed

I was having the same problem in Linux Ubuntu 18. After the update from Ubuntu 17.10, every git command would show that message.

The way to solve it is to make sure that you have the correct permission on the id_rsa and id_rsa.pub.

Check the current chmod number by using stat --format '%a' . It should be 600 for id_rsa and 644 for id_rsa.pub.

To change the permission on the files use

chmod 600 id_rsa
chmod 644 id_rsa.pub

That solved my issue with the update.

There could be various reason for getting the SSH error:

sign_and_send_pubkey: signing failed: agent refused operation

Some of them could be related to the issues highlighted by the other answers (see this thread answers), some of them could be hidden and thus would require a closer investigation.

In my case I've got the following error message:

sign_and_send_pubkey: signing failed: agent refused operation

user@website.domain.com: Permission denied (publickey,gssapi-keyex,gssapi-with-mic)

The only way to find the real problem was to invoke the -v verbose option which resulted in printing a lot of debugging info:

debug1: Connecting to website.domain.com [xxx.xxx.xxx.xxx] port 22.
debug1: Connection established.
debug1: identity file /home/user/.ssh/id_rsa.website.domain.com type 0
debug1: key_load_public: No such file or directory
debug1: identity file /home/user/.ssh/id_rsa.website.domain.com-cert type -1

Please note that the line saying key_load_public: No such file or directory is referring the next line and not the previous line.

So what SSH really says is that it could not find the public key file named id_rsa.website.domain.com-cert and that seemed to be the problem in my case since my public key file did not contain the -cert suffix.

Long story short: the fix in my case was just to make sure that the public key file was named as expected. I could never suspected that without debugging the connection.

The bottom line is USE THE SSH VERBOSE MODE (-v option) to figure out what is wrong, there could be various reasons, none that could be found on this/another thread.

quick note for those recently upgrading to "modern" ssh version [OpenSSH_8.1p1, OpenSSL 1.1.1d FIPS 10 Sep 2019] - supplied with fedora 31, seems not to be anymore accepting old DSA SHA256 keys (mine are dated 2006!) - created a new rsa key, public added to authorized, private on client, and everything works perfectly.

thanks for previous suggestions, especially the ssh -v has been very useful

Run the below command to resolve this issue.

It worked for me.

chmod 600 ~/.ssh/id_rsa

What worked here : on the client

1) ssh-add

2) ssh-copy-id user@server

The keys has been created some time ago with plain "ssh-keygen -t rsa" I sw the error message because I copied across my ssh public key from client to server (with ssh-id-copy) without running ssh-add first, since I erroneously assumed I'd added them some time earlier.

I had the error when using gpg-agent as my ssh-agent and using a gpg subkey as my ssh key https://wiki.archlinux.org/index.php/GnuPG#gpg-agent.

I suspect that the problem was caused by having an invalid pin entry tty for gpg caused by my sleep+lock command used in my sway config

bindsym $mod+Shift+l exec "sh -c 'gpg-connect-agent reloadagent /bye>/dev/null; systemctl suspend; swaylock'"

or just the sleep/suspend

Reset the pin entry tty to fix the problem

gpg-connect-agent updatestartuptty /bye > /dev/null

and the fix for my sway sleep+lock command:

bindsym $mod+Shift+l exec "sh -c 'gpg-connect-agent reloadagent /bye>/dev/null; systemctl suspend; swaylock; gpg-connect-agent updatestartuptty /bye > /dev/null'"

I once had a problem just like yours, and this is how I solved it through the following steps.

  1. chmod 700 ~/.ssh
  2. chmod 600 ~/.ssh/*
  3. ssh-copy-id user@ip
  4. ssh-agent -s
  5. ssh-add

First ssh-add then ssh user@ip

this worked for me

As others have mentioned, there can be multiple reasons for this error.

If you are using SSH with Smart Card (PIV), and adding the card to ssh-agent with
ssh-add -s /usr/lib64/pkcs11/opensc-pkcs11.so
you may get the error
sign_and_send_pubkey: signing failed: agent refused operation
from ssh if the PIV authentication has expired, or if you have removed and reinserted the PIV card.

In that case, if you try to do another ssh-add -s you will still get an error:
Could not add card "/usr/lib64/opensc-pkcs11.so": agent refused operation

According to RedHat Bug 1609055 - pkcs11 support in agent is clunky, you instead need to do

ssh-add -e /usr/lib64/opensc-pkcs11.so
ssh-add -s /usr/lib64/opensc-pkcs11.so

Yes. Run ssh-add on the client machine. Then repeat command ssh-copy-id userserver@012.345.67.89

For me the problem was a wrong copy/paste of the public key into Gitlab. The copy generated an extra return. Make sure what you paste is a one-line key.

In my case the problem was that GNOME keyring was holding an invalid passphrase for the ssh key to be used. After spending indecent amount of time troubleshooting this issue I ran seahorse and found the entry to hold empty string. I can only guess that it was caused by mistyping the passphrase at first use some time earlier, and then probably cancelling the requester or so in order to fall back to command line. Updating the entry with correct passphrase immediately solved the problem. Deleting that entry (from "login" keyring) and reentering passphrase at that first prompt (and checking the appropriate checkbox) solves this too. Now agent gets the correct passphrase from the unlocked at login keyring named "login" and neither asks for passphrase nor "refuses operation" anymore. Of course YMMV.

To this error:

# git pull
sign_and_send_pubkey: signing failed: agent refused operation
git@github.com: Permission denied (publickey).    
fatal: Could not read from remote repository.

Please make sure you have the correct access rights and the repository exists.

Verify or add again the public key in Github account > profile > ssh.

I solved like this:

# chmod 400 ~/.ssh/id_rsa

# ls  ~/.ssh/id_rsa -ls  
4 -r--------. 1 reinaldo reinaldo 1679 Jul 26  2017 /home/reinaldo/.ssh/id_rsa

# git pull                                 
remote: Counting objects: 35, done.
remote: Compressing objects: 100% (19/19), done.
remote: Total 35 (delta 9), reused 34 (delta 9), pack-reused 0
Unpacking objects: 100% (35/35), done.

Thank you.

This should be rather a SuperUser question.

Right I have the exact same error inside MacOSX SourceTree, however, inside a iTerm2 terminal, things work just dandy.

However, the problem seemed to be that I've got two ssh-agents running ;(

The first being /usr/bin/ssh-agent (aka MacOSX's) and then also the HomeBrew installed /usr/local/bin/ssh-agent running.

Firing up a terminal from SourceTree, allowed me to see the differences in SSH_AUTH_SOCK, using lsof I found the two different ssh-agents and then I was able to load the keys (using ssh-add) into the system's default ssh-agent (ie. /usr/bin/ssh-agent), SourceTree was working again.

After upgrading Fedora 26 to 28 I faced same issue. And following logs were missing

/var/log/secure
/var/log/messages

ISSUE:

antop@localmachine  ~  ssh root@ocp1.example.com
sign_and_send_pubkey: signing failed: agent refused operation
root@ocp1.example.com's password:

error message is not pointing actual issue. Issue resolved by

chmod 700 ~/.ssh
chmod 600 ~/.ssh/*

I need to share, as I spent too much time looking for a solution

Here was the solution : https://unix.stackexchange.com/a/351742/215375

I was using this command :

ssh-keygen -o -t rsa -b 4096 -C "email@example.com"

gnome-keyring does not support the generated key.

Removing the -o argument solved the problem.

2022 WebDevInsider