[ISSUE] Entropy very different on physical vs virtual machines

[bcookatpcsd <bcookatpcsd@users.noreply.github.com>]

[-- Attachment #1: Type: text/plain, Size: 2799 bytes --]

New issue by bcookatpcsd on void-packages repository

https://github.com/void-linux/void-packages/issues/29510

Description:
void running in esxi w/ open-vm-tools

`Linux Void69 5.10.23_1 #1 SMP 1615515203 x86_64 GNU/Linux
[I] bcook@Void69 ~> cat /proc/sys/kernel/random/entropy_avail
3420
`
` cat /etc/sv/vmtoolsd/run 
#!/bin/sh

set -e

/sbin/modprobe vmw_vsock_vmci_transport

exec vmtoolsd

`

void running on bare metal machine (not running intended services yet)

`[I] root@VoidClinton ~# uname -a
Linux VoidClinton 5.10.23_1 #1 SMP 1615515203 x86_64 GNU/Linux
[I] root@VoidClinton ~# cat /proc/sys/kernel/random/entropy_avail
7
`

another void physical machine running docker and three containers (dns processes) AND chrony

`[I] bcook@Void63 ~> uname -a
Linux Void63 5.10.23_1 #1 SMP 1615515203 x86_64 GNU/Linux
[I] bcook@Void63 ~> cat /proc/sys/kernel/random/entropy_avail
2869
`

`[I] bcook@Void63 ~ [1]>  chronyc sources 
MS Name/IP address         Stratum Poll Reach LastRx Last sample               
===============================================================================
^+ time1.google.com              1  10   377   823  -1211us[-1292us] +/-   12ms
^* time4.google.com              1  10   377   612   -996us[-1078us] +/- 9027us
^+ time3.google.com              1  10   377   768   -169us[ -251us] +/- 8996us
^+ time2.google.com              1  10   377   756   +151us[  +69us] +/-   12ms
^+ time.cloudflare.com           3  10   377   256   -634us[ -634us] +/- 7338us
^+ time.cloudflare.com           3  10   377   292   -218us[ -218us] +/- 7610us
`

Where is the entropy coming from? should I run rng-tools or haveged.. 

What about irqbalance? github shows work still being done there.. 

via glances:

![image](https://user-images.githubusercontent.com/55087301/111325871-7c853400-8642-11eb-870d-1ae6bc66dcbe.png)

[I] bcook@Void63 ~> cat /proc/interrupts | grep LOC 
LOC:    7904516    7711579   10060743    7701411   Local timer interrupts
[I] bcook@Void63 ~> uptime 
 10:30:21 up 18:27,  0 users,  load average: 0.13, 0.15, 0.06


I thought LOC had to do with proper interrupt usage (affinity being correctly used) vs an actual "interrupted performance" counter.. 

![image](https://user-images.githubusercontent.com/55087301/111326540-02a17a80-8643-11eb-82c1-e548e1628304.png)


the local eno1 interface seems pinned to CPU2, hence why the LOC counters are higher for CPU2.. 

rng-tools,haveged,irqbalance.. are those still used in modern systems? Are those not in the kernel? Are they recommended in 2021? 

I also have TZ='America\New_York' set in /etc/profile.. 
https://blog.packagecloud.io/eng/2017/02/21/set-environment-variable-save-thousands-of-system-calls/

Thanks in advance..