Cannot update config with cmdeploy

Dio · February 21, 2026, 9:04am

Hi,

I’ve got a problem with my chatmail test relay. I just wanted to update some var in the config and run cmdeploy run again but it fails:

~/relay 1.9.0 ❱ scripts/cmdeploy run --verbose
[ssh] login to d9.example.com
[ssh] Collecting initial DNS settings.
.............[$ pyinfra --ssh-user root d9.example.com /Users/myuser/relay/cmdeploy/src/cmdeploy/run.py -y]
--> Loading config...
--> Loading inventory...
--> Connecting to hosts...
    [d9.example.com] Could not connect (timed out)

--> Disconnecting from hosts...
--> pyinfra error: No hosts remaining!
Deploy failed

Connection via ssh works and I can see cmdeploy connecting:

Feb 21 09:03:11 d9 sshd[8483]: Accepted publickey for root from ...
Feb 21 09:03:11 d9 sshd[8483]: pam_unix(sshd:session): session opened for user root(uid=0) by (uid=0)
Feb 21 09:03:11 d9 sshd[8483]: pam_env(sshd:session): deprecated reading of user environment enabled

Any ideas?

link2xt · February 21, 2026, 12:04pm

Try removing venv and recreating it with scripts/initenv.sh.

Dio · February 21, 2026, 8:40pm

Same result afterwards.

link2xt · February 21, 2026, 10:14pm

Do you have anything special configured in ~/.ssh/config for this host?

It is very likely a pyinfra issue, there is a similar bug here:

github.com/pyinfra-dev/pyinfra

Pyinfra tries to connect to first ip only.

opened 02:25PM - 16 Mar 22 UTC

romaia

bug dependency issue

## Describe the bug I have a situation my hosts have multiple IPs, but only one… is accessible. Pyinfra seems to only try to connect to the first ip resolved. For instance: ```$ host myserver myserver has address 172.16.0.4 myserver has address 10.100.48.54 ``` If (by any chance) the order is reversed, pyinfra manages to connect. SSH also manages to connect: it will timeout for the first ip, but will try the second after that. I have also decreased `ConnectTimeout` in my ssh config to speed up the first timeout. ## To Reproduce Make one host have multiple ips, one being invalid or inacessible. ## Expected behavior pyinfra should try to connect to the second ip after the first timeout. ## Meta + Include output of `pyinfra --support`. ``` System: Linux Platform: Linux-5.4.0-48-generic-x86_64-with-glibc2.29 Release: 5.4.0-48-generic Machine: x86_64 pyinfra: v1.6.3 Executable: /home/stoq_pdv_user/.local/bin/pyinfra Python: 3.8.5 (CPython, GCC 9.3.0) ``` + How was pyinfra installed (source/pip)? pip + Consider including output with `-vv` and `--debug`. ``` $ pyinfra wlx138pdv070 fact facts.InstalledVersion -vvv --debug [pyinfra_cli.main] Deploy directory remains as cwd --> Loading config... --> Loading inventory... [pyinfra_cli.inventory] Creating fake inventory... --> Connecting to hosts... [pyinfra.api.connectors.ssh] Connecting to: wlx138pdv070 ({'allow_agent': True, 'look_for_keys': True, 'hostname': 'wlx138pdv070', '_pyinfra_force_forward_agent': None, '_pyinfra_ssh_config_file': None, 'timeout': 10}) [wlx138pdv070] Could not connect (timed out) [pyinfra.api.state] Failing hosts: wlx138pdv070 --> pyinfra error: No hosts remaining! ```

Does your host have multiple addresses, e.g. IPv4 and IPv6? Could it be that only one is actually responding on port 22 and the other is filtered out?

Dio · February 21, 2026, 10:52pm

Bingo! The IPv6 DNS entry was missing a single digit. While setting up the test server I did not have IPv6 connectivity at all, so this was no harm and did not get noticed.

Thank you!