Antrea agent pod fail to start due to "yaml: unmarshal errors"
Article ID: 409365
Updated On:
Products
Issue/Introduction
Antrea agent pod fail to start due to error in log fetched by `kubectl logs antrea-agent -n kube-system`,
I0902 08:58:39.460220 1 log_file.go:93] Set log file max size to 104857600
F0902 08:58:39.461170 1 main.go:47] Failed to complete: yaml: unmarshal errors:
line 184: field packetInRate not found in type agent.AgentConfig
line 324: field enable not found in type agent.AntreaProxyConfig
line 390: field auditLogging not found in type agent.AgentConfig
Cause
The problem is directly caused by version mismatch between antrea-agent deployment yaml and image file. For example, in the case that image file is at v1.13 but deployment yaml is at v2.1, antrea-agent process will fail to start because it expects properties which do not exist in configuration files.
In TKGI, the possible trigger is - a cluster upgrade did not complete successfully, it enters into a state - deployment files are updated, but pod images in worker cache get rolled back after worker recreation by `bosh recreate` operation.
Resolution
To resolve the problem, please re-run upgrade with the impacted TKGI cluster, so that pod images will match deployment version.
Feedback
