Unable to ssh to Lightsail server when in burst zone with ample remaining burst capacity

Question

Hi,
I have noticed that I can't access lightsail server when it is in burst zone. It doesn't matter if I have a lot of burst capacity remaining. I tried this by creating a new server etc.. same issue. Here's how to replicate the issue:

1. Create a Lightsail server (in my case it's a plan with 2vCPUs and 30% sustainable capacity)
2. Wait for remaining burst capacity to reach 100% (~432m)
3. Run any script that will consume bust CPU (I was running it at 50% instead of the sustainable 30%, my guess is 1vCPU was maxed out in my case)

After a few mins (maybe give it 30-40 mins to be sure) you will not be able to ssh (even the "connect via ssh button on AWS interface won't work). This should not happen as from my understanding of how burst capacities work I should be able to burst for much longer than that. The main problem here is because I can't ssh, I can't stop the script either. My main use-case for these bursts is to be able to run scripts for updates/upgrades temporarily. But this weird behavior all but makes Lightsail entirely useless as a server.

Answer

> As I understand, whenever you run a script consuming CPU/Memory on your LightSail server, you are not being able to SSH into the server, even when Burstable capacity is not entirely depleted.

___

First, please allow me to explain Burst performance baselines:

Performance baselines are per vCPU. The CPU utilization metric graph in the Lightsail console averages the CPU utilization and baseline for instances with more than one vCPU. For example, a Linux or Unix-based $40 USD/month instance has two vCPUs and an averaged CPU utilization baseline of 30%.

- https://lightsail.aws.amazon.com/ls/docs/en_us/articles/amazon-lightsail-viewing-instance-burst-capacity

Therefore, if:

- One vCPU operates at 50% and the other at 0%, a 25% averaged CPU utilization is displayed on the graph. This puts the instance's CPU utilization below its 30% baseline, and in the sustainable zone.

- One vCPU operates at 30%, and the other at 20%, a 25% averaged CPU utilization is displayed on the graph. This puts the instance's CPU utilization below its 30% baseline, and in the sustainable zone.

- One vCPU operates at 35% and the other at 25%, a 30% averaged CPU utilization is displayed on the graph. This puts the instance's CPU utilization at the 30% baseline.

- One vCPU operates at 100% and the other at 90%, a 95% averaged CPU utilization is displayed on the graph. This puts the instance's CPU utilization above its 30% baseline, and in the burstable zone.

___

In order to understand per vCPU performance, we would need to investigate what is causing your instance to burst using tools like top on Linux/Unix instances, and Task Manager on Windows Server instances. In top, we can also check per vCPU consumption, determining the Burst baseline performance.

___

Next, there could be multiple reasons behind server becoming unreachable that we would need to dive deeper into. You would need need to share more details if you want help with this issue - Relevant log snippets with errors, what's happening when the instance crashes, any troubleshooting performed so far, exact configuration of the instance (since there are 2 LightSail instances offered with 2 vCPUs, and depending on the configuration, burst time will vary, etc. When you run the script, we would also need to check the Memory metrics from within the instance using utility like 'top', in order to understand whether memory is not being depleted.

___

Lastly, I would recommend opening up a case with us at AWS Premium Support for an AWS Engineer to check instance metrics live on screenshare when you run the script. Therefore, please open up a case with us and choose the Chat/Call option in order to investigate further.

Unable to ssh to Lightsail server when in burst zone with ample remaining burst capacity

相关内容