we are running a 3-node Kubernets cluster v1.28.10 with Cilium v1.15.5 CNI which provides various applications for the surrounding network, e.g. a log server based on syslog-ng.
It is also used by the Kubernetes nodes themselves as a remote syslog server. We noticed that all nodes appear as clients in the log server with their main ip address (eth0), except for the node on which the syslog-ng pod/entpoint is running. This node appears with the ip address of its cilium_net adapter, which is not, what we expect. As a result, the log destination, for this node is wrong as it is based on the client hostname (dns reverse lookup of client ip).
Is there any option to connect the "in-cluster" service from all nodes using there main ip address (eth0) as source address?
Is there an existing issue for this?
What happened?
Hello team,
we are running a 3-node Kubernets cluster v1.28.10 with Cilium v1.15.5 CNI which provides various applications for the surrounding network, e.g. a log server based on syslog-ng.
It is also used by the Kubernetes nodes themselves as a remote syslog server. We noticed that all nodes appear as clients in the log server with their main ip address (eth0), except for the node on which the syslog-ng pod/entpoint is running. This node appears with the ip address of its cilium_net adapter, which is not, what we expect. As a result, the log destination, for this node is wrong as it is based on the client hostname (dns reverse lookup of client ip).
Is there any option to connect the "in-cluster" service from all nodes using there main ip address (eth0) as source address?
Thanks and best regards, Don.
Cilium ConfigMap looks as follows:
Cilium Version
1.15.5
Kernel Version
5.14.21-150500.55.59-default
Kubernetes Version
1.28.10
Regression
No response
Sysdump
No response
Relevant log output
No response
Anything else?
No response
Cilium Users Document
Code of Conduct