Recently an AWS account I'm handling has seen a large increase in the Regional Data Transfer cost. Inspecting the usage / bill details reveals this is down to an increase in "ELB data" being processed and "Regional Data Transfer".
All the instances are in the same availability zone, so the issue must be the load balancers. There are 2 ELBs for internet-facing traffic and 2 ELBs for internal traffic, however looking at the console I can see all 4 are Classic ELBs talking to instances inside EC2-Classic rather than a VPC.
Because of this, the 2 "internal" ELBs are being referenced by a DNS name in the form
ELB-Name-loadbalancer-xxxxxxx.eu-west-1.elb.amazonaws.com, rather than the
Internal-loadbalancer-xxxxxxx.eu-west-1.elb.amazonaws.com created with internal ELBs.
Presumably these DNS records resolve to public IPs, and would therefore incur the data-transfer cost? If this is the case, variations in the TTL could theoretically cause the cost variations I'm seeing.
Can anyone confirm if this is correct or if I'm mistaken (and if so, what else I could do to help trace the problem)?
EDIT: Here's a chart of what I'm seeing:
While data in/out is flat, ELB data & Regional transfer are increasing like crazy. Interestingly ELB data is roughly half of the regional data transfer, so I think that would correlate with my guess that the ELB traffic is getting charged in/out. Unfortunately I have no idea what's causing the actual increase in the ELB data itself.