1

Student here, I'm wondering how we know that hosts are degrading (specifically EC2 instances)? What are some typical behaviors that could indicate the degradation? I looked through the site and googled, and the best answer that I found is Why do AWS ec2 instances get degraded? and the AWS EC2 website is also not much of much help

What else is there to watch out for? Resources would be greatly appreciated!

From the other post, I know "high I/O, memory pressure, high swap use" could be causes, and any more detailed explanation for those would also be nice. Thanks in advance!

Edit: I was looking for sources like https://www.datadoghq.com/blog/top-5-ways-to-improve-your-aws-ec2-performance/ and https://www.splunk.com/en_us/blog/devops/12-top-things-to-monitor-in-amazon-ec2.html. Any more resources would be helpful!

myao
  • 13
  • 3
  • 2
    What do you mean by “degrading”? - Running a crap application in the cloud won’t make it smell like roses (and won’t magically fix memory leaks and such) and when you treat virtual compute capacity like the dinosaurs of your datacenter and leave it running for a decade or more obviously you will get the same kind of issues but otherwise?...? – Bob Feb 16 '21 at 23:07
  • Ah, I meant this as a more general question outside the scope of the software. Just in general, what are good indicators that a host needs restarting? – myao Feb 16 '21 at 23:29
  • EC2 status checks, Cloudwatch Metrics like CPU / RAM trending up, application specific metrics. https://docs.aws.amazon.com/AWSEC2/latest/UserGuide/monitoring-system-instance-status-check.html – Tim Feb 17 '21 at 00:02

1 Answers1

2

As far as I know most operating systems nowadays are actually quite good and stable but when you are regularly applying security updates and bug fixes you will need to reboot every once in while as well. Do that and you will avoid many issues.

When the hosting provider is over commissioning cpu (in general but particularly when you are impacted on the specific hypervisor where your VPS is running ) you will see that as high steal time in the guest. Detecting high steal time is a good moment to recycle a VPS (destroy outright or do virtual power off and power on again hoping that your VPS will be restarted on a different hypervisor).

Bob
  • 5,335
  • 5
  • 24
  • Thank you! I will look into steal time, since I'm not familiar with infrastructure this is very helpful. – myao Feb 16 '21 at 23:58