Mirantis / virtlet

Kubernetes CRI implementation for running VM workloads
Apache License 2.0
743 stars 128 forks source link

Virtlet lack of entropy #843

Open roboman2444 opened 5 years ago

roboman2444 commented 5 years ago

My Virtlet pods have a surprising lack of entropy, and therefore, fail/timeout some services that require /dev/random and getrandom().

pod after being up for a few minutes

It appears that something like https://wiki.qemu.org/Features/VirtIORNG would help. Is there any way to enable this in Virtlet, or some other method of generating more entropy on the pods?

The same image generates entropy just fine when running on my desktop in Qemu.

My issue might be related to the ec2 i3.metal instance I am running it on, but that has > 3000 in /proc/sys/kernel/random/entropy_avail.

I'm running images based on Alpine if that matters. (generated with https://github.com/alpinelinux/alpine-make-vm-image), and the Host is running Ubuntu 18.10 with kubernetes 1.11.6.

I'm currently using haveged as a band-aid fix for this.

stale[bot] commented 5 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

stale[bot] commented 5 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days if no further activity occurs. Thank you for your contributions.

jellonek commented 5 years ago

@ivan4th @pigmej can you reopen that and teach somehow stale bot about to skip this one? ;)

pigmej commented 5 years ago

Done :)