-
Bug
-
Resolution: Done
-
Normal
-
4.18
-
None
-
False
-
-
None
-
Moderate
-
None
-
None
-
None
-
Agent Sprint 282
-
1
-
Done
-
Bug Fix
-
In baremetal installations with VLANs, it is possible for the agent-installer pre-network-manager-config service to timeout. This fix increased the timeout to handle these type of installations.
-
None
-
None
-
None
-
None
This is a clone of issue OCPBUGS-63386. The following is the description of the original issue:
—
Description of problem:
During agent-based OpenShift cluster installations on bare metal environments with VLAN interfaces, CoreOS fails to bootstrap networking automatically. The pre-network-manager-config.service times out during execution, preventing NetworkManager from applying configuration profiles. As a result, the node boots without any active network interfaces, and installation cannot proceed unless networking is manually configured through nmcli or nmtui via console access.
Journal logs from affected nodes show:
pre-network-manager-config.service: start operation timed out. Terminating. ls: cannot access '/tmp/tmp.../hostX/*.nmconnection': No such file or directory
Manually increasing the timeout (TimeoutStartSec=300s) in the systemd unit file allows the service to complete successfully, which confirms that the service times out before completing its configuration tasks. This issue affects automation of agent-based installations and should be fixed permanently—either by increasing the default timeout value or by making it configurable.
Impact:
Automation of agent-based OpenShift cluster installation on bare metal nodes is blocked due to networking not being initialized automatically.
Workaround:
Patch the ignition or modify the pre-network-manager-config.service to increase TimeoutStartSec to 300 seconds, then reload and restart the service manually.
Expected results:
During agent-based installation on bare metal systems, the CoreOS node should automatically configure and bring up networking without any manual intervention.
Additional info: