1 個回答
- 最新
- 最多得票
- 最多評論
0
Where are you seeing the limit of ~ 55,000 connections?
I think you're reading this page in the documentation and I'd note that it is 55,000 connections per target (so ECS container in this case); and it applies when source IP address preservation is enabled (which is the default).
Ah yes, that was was I was looking at, so clearly the single NLB will continue to grow to accomodate!
Thank you for clarifying that!
相關內容
- AWS 官方已更新 1 年前
- AWS 官方已更新 2 年前
Can you share more information on your setup. Is this a single AZ NLB with a single IP target that is hovering around 52k active flows?
Thanks and sure! It's all in us-east-1 and there are 4AZ's that NLB is serving. I am not sure what you mean by a single IP target, but these are all MQTT connections that are made to a server farm of 15 or so machines using one DNS name .. There are a few listeners on the NLB but the main one is 99% of the traffic and as I said used for MQTT connections.
I also updated the question to include a diagram of the current and what I need for the updated (how I would picture it) in the original question.
When I look at the NLB / monitoring I see that Active Flow Count sitting at 51k or so.
I do want to start to look at moving to a second Location like us-west-1, but right now all traffic from the outside hardware devices goto mqtt.example.com. That points to that single NLB. So to combat that 55k, I would say for now, it would be great to start a 2nd NLB, mirror things, goto Route53 and say mqtt.example.com goes to CNAME NLB1, NLB2 and have that split the traffic but that is not possible. So this is where I am trying to figure out how I split that if I can't simply update the remote hardware to say some goto mqtt.example.com and others goto mqtt2.example.com or something like that.
As I said, once I get the concept and best practice, I am sure it will work as I start to roll out more listening devices to the us-west, etc. however this is something I clearly need to understand a bit more and have solved sooner than later.