Recently we had an issue with our EC2 instance where some packets were intermittently dropped. The instance had two NICs with IPs on the same subnet, and the issue was traced (with the help of AWS support) to the lack of ip rule for primary NIC. Because the two NICs were on the same subnets under rare conditions the returning packets for NIC1 were sent out via NIC2 which were dropped.
While the behavior is documented (
|
# We only create rules for secondary interfaces so |
), we would appreciate it if you could also document the implications of the missing rule in the scenario where the instance has one or more secondary NICs with IPs that share the same subnet as the primary NIC.
Thanks for your considerations.
Recently we had an issue with our EC2 instance where some packets were intermittently dropped. The instance had two NICs with IPs on the same subnet, and the issue was traced (with the help of AWS support) to the lack of ip rule for primary NIC. Because the two NICs were on the same subnets under rare conditions the returning packets for NIC1 were sent out via NIC2 which were dropped.
While the behavior is documented (
amazon-ec2-net-utils/lib/lib.sh
Line 711 in b1a7565
Thanks for your considerations.