I launched two identical instances of the middle processor on Amazon for a load balancer for several months. I noticed that the load balancer has the habit of declaring an instance unhealthy on a fairly regular basis, dropping the instance and replacing a new instance of a specific AMI.
This is technically the right thing, I just don’t understand why she thinks that the case is unhealthy, occasionally. I have been monitoring health check ports for the last 3 days, and checking every 60 seconds works constantly when using the public DNS of two instances. The load balancer declared 3 times unhealthy during this period and replaced it. Instances are massively suppressed for what I need, purposefully, so I can eliminate this because of a problem.
With ELB architecture, I know that it doesn't really matter, but the level of ill health has gone from once a week to more than once a day. Each instance that is twisted requires an additional hour of instance cost from me. If it gets worse, the cost will become non-trivial, but more importantly, it does not give me faith in the internal organs of ELB.
This is not the question that this , my random failure. For information, I use the EU / Ireland data center, and my unhealthy criterion is 10 failures in my port (8080) in a 5-minute period (which is longer than I really would like to install anyway, I don’t want traffic going to copies that do not receive a response within 5 minutes).
I know that someone is going to offer to contact Amazon, but I do not have a support contract, and anyone who has tried this knows what answer I will get if I get it at all. I really like the idea of this thing, it just does not seem stable to me.
source share