You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently there is a limitation on being able to use Nat gateways with resiliency i.e. across multiple AZ's. The AWS architecture is:
Compute Subnet (eu-west-1 AZ A) -> RT_A -> NG_A -> DF_RT -> IGW
Compute Subnet (eu-west-1 AZ B) -> RT_B -> NG_B -> DF_RT -> IGW
Compute Subnet (eu-west-1 AZ C) -> RT_C -> NG_C -> DF_RT -> IGW
RT = Route table
NG = Nat gateway
DF = Default
IGW = Internet Gateway
A to C = is the availability zones
The issue is that flannel at the moment only supports adding the route tables to a specified route table
cfg := struct {
RouteTableID string
}{}
At the moment this is a string if undefined it finds the route table associated with the instance and it will add it there. In this instance, it means that routes will only be populated in one route table i.e. RT_A or RT_B etc, hence any traffic from RT_B to RT_A won't know how to route, we would therefore need to populate it as an array of route table ID's.
I am happy to do a PR for this. Without this, traffic can't route accordingly when you are using NAT gateways or decide to use more than one route table for your cluster (for whatever reason).
The text was updated successfully, but these errors were encountered:
Currently there is a limitation on being able to use Nat gateways with resiliency i.e. across multiple AZ's. The AWS architecture is:
Compute Subnet (eu-west-1 AZ A) -> RT_A -> NG_A -> DF_RT -> IGW
Compute Subnet (eu-west-1 AZ B) -> RT_B -> NG_B -> DF_RT -> IGW
Compute Subnet (eu-west-1 AZ C) -> RT_C -> NG_C -> DF_RT -> IGW
RT = Route table
NG = Nat gateway
DF = Default
IGW = Internet Gateway
A to C = is the availability zones
The issue is that flannel at the moment only supports adding the route tables to a specified route table
At the moment this is a string if undefined it finds the route table associated with the instance and it will add it there. In this instance, it means that routes will only be populated in one route table i.e. RT_A or RT_B etc, hence any traffic from RT_B to RT_A won't know how to route, we would therefore need to populate it as an array of route table ID's.
I am happy to do a PR for this. Without this, traffic can't route accordingly when you are using NAT gateways or decide to use more than one route table for your cluster (for whatever reason).
The text was updated successfully, but these errors were encountered: