Description of problem:

On OpenShift clusters where the administrators choose to use non-private networks (RFC 1918 and RFC 6598) for the cluster pod network, it is not possible to deploy LokiStack cluster. Due to the default assignment in the hash ring configuration to advertise IP address from private IP networks only, the ingester, distributor and querier component are able to form a healthy gossip ring.

Version-Release number of selected component (if applicable):

OpenShift Logging 5.5

OpenShift Logging 5.6

How reproducible:

Always

Steps to Reproduce:

1.  
2.  
3. ...

Actual results:

1. Crashlooping pods:

logging-loki-compactor-0                      0/1     CrashLoopBackOff   
logging-loki-distributor-74f676459-9gbzh      0/1     CrashLoopBackOff   
logging-loki-distributor-74f676459-zlqhj      0/1     CrashLoopBackOff   
logging-loki-ingester-0                       0/1     CrashLoopBackOff   
logging-loki-querier-6d56f558b-rnflv          0/1     CrashLoopBackOff   
logging-loki-querier-6d56f558b-z7h8f          0/1     CrashLoopBackOff

1. Services from pods has no endpoints.
2. Error message:

failed to create memberlist: Failed to get final advertise address: no private IP address found, and explicit IP not provided 

Expected results:

All LokiStack pods required to form a health hash ring advertise IP address from the pod network range. In turn all pods are healthy and normal operations resume.

Additional info:

KCS: failed to create memberlist no private IP address found" in Loki for OCP 4