When deploying code from GitLab with Code Manager in Puppet Enterprise or Continuous Delivery for PE, it fails with an SSH error.
Error messages and logs
Failed to start SSH session: Failed getting banner at /opt/puppetlabs/server/data/code-manager/worker-caches/deploy-pool-2/ssh---git@git.example.com-control-repo.git
GitLab uses the OS sshd service, which is pre-configured for human users, to handle incoming traffic. Ten is a common OS default for the number of SSH connections allowed to start at any time. When SSH is used to serve Git repositories, the number of concurrent connections commonly exceeds 10, so connections might be denied.
Code Manager defaults to deploying up to two environments simultaneously via deploy_pool_size
, and each environment can run up to four simultaneous git clone operations via download_pool_size
, using eight connections (80% of capacity). If other services or users are using an untuned GitLab installation alongside Code Manager, operations will fail intermittently.
Version and installation information
PE version: All supported versions
Continuous Delivery for PE version: All supported versions
Solution
Note: We cannot troubleshoot third-party software. For assistance with GitLab, see Get help with GitLab.
To fix the issue:
-
Increase the number of concurrent SSH connections by configuring
MaxStartups
. -
Restart the sshd daemon.
How can we improve this article?
0 comments
Please sign in to leave a comment.
Related articles