1

I've been experimenting with enabling hibernation on m5a.2xlarge machines (32G RAM, 8vCPU, up to 2880MB/s EBS bandwidth), and my observation is that the machines fairly consistently take upwards of 20-30 minutes in the 'stopping' state. Trying to understand if this is expected (I'd hope not?), or whether there are any recommendations for how I could debug the behavior.

I have given the machines an extra 32G on their root volume to accommodate the additional storage requirements of hibernation, and encrypted the disk.

Seth Nelson
  • 11
  • 1
  • What problem is this causing? Do you need to restart them again quickly? Or are you just curious? You only pay when the instance is in the running state ( https://aws.amazon.com/premiumsupport/knowledge-center/ec2-instance-hour-billing/ ). 30 mins seems quite slow, but I've never used hibernate - I just stop the instances. – Tim Jul 12 '22 at 08:33
  • I am using this in the context of a developer tool - while my goal is that we never hibernate a machine that is in-use or will soon be in-use, it may occasionally happen, and in those cases speed in a concern. But also, it just didn't pass the sniff test, so I was hoping to discover a quick win. – Seth Nelson Jul 13 '22 at 19:36
  • Why don't you just stop the machine? Stop and start happens a lot faster than hibernate by the sounds of things – Tim Jul 13 '22 at 19:53
  • By that logic, why did Amazon introduce Hibernate functionality at all? Because resuming memory state is valuable in a number of circumstances, e.g. expensive program setup. – Seth Nelson Jul 14 '22 at 21:57
  • Good question. I haven't looked at it myself so I don't know. Hibernate should take seconds I would think, not minutes. Reach out to AWS Support if you have access, they're always good. – Tim Jul 14 '22 at 22:07

0 Answers0