- 最新
- 投票最多
- 评论最多
In theory, given the 16 bit number that represents the source port number; and subtracting for "reserved" ports on any particular operating system, ~60k sessions from a single IP address is about right. Especially if it is a single port that the application is connecting to.
The reason you're seeing far more than that to ALB directly is probably because your connections are happening to each ALB node. If you look at the DNS response for the ALB you should see two or three IP addresses; connections to those will multiply the ~60k number.
So this isn't an AWS limit; it's a TCP (and UDP) limit.
Thank you for the quick response. I found that limit, but because the Global Accelerator has two static IPs, I was expecting around 120k before starting to see the cap. The "reserved" ports may bring the count down a bit more, or one IP may be preferred above another one cause of the location of the source IP.
Are you sure you're connecting to both Global Accelerator IP addresses?
While AWS Global Accelerator provides two active Anycast IP addresses for redundancy and high availability, it's normal and expected for individual clients to consistently use just one of these addresses for their connections. The apparent use of a single IP doesn't mean the second IP isn't active or available; it's simply a result of typical client behavior and connection management. Please check your clients application configurations too.