Issue with CPU load reported on Guest Vs ESXI Host
-
Hi all,
There is a certain disparity between my CPU load reported on the Guest and ESXI host.
For example right now the Firewall is sitting there doing nothing with about 1% load.
Yet in ESXI it's using about 1ghz cpu!?
Something I missed here? I am using the vmware tools so wouldn't expect this to happen.
Thanks,
Josh
-
53 views and no thoughts?
-
Having the same issue here, pfsense is showing between 2 and 5% but ESXi is reporting 1 GHz to 2 GHz.
I have installed the ESXi official tools but using E1000 nic's
Testing i get a full 1gb so the "issue" is not causing direct problems. I also had a openvpn site2site running over 100mbit wlan.So far i have not found a solustion, running no tools, official tools or open tools makes no difference
-
Here are my screenshots showing high cpu if pfsense is handeling traffic.
Also i have tried soms stuff in this post : http://forum.pfsense.org/index.php?topic=41647.0
The issue there seems to be the same.
-
glad it's not just me then!
I have no idea what is taking up the CPU :(
-
Well, for me it will not impact performance, so it seems to be ESXi not reporting it correct.
If ESXi is at 100% you would experct te hit a bandthwidth limit in pfsense, by for me that is not the case.We are using multiple vlan's on one nic.
-
Noebas and Deadringers,
What versions of pfSense and ESXi are you running?
How many vCPUs did you allocate to pfSense?
-
I have Pfsense 2.0.3 running om ESXi 5, build xxxx this is a cloud server so build is unknow.
1cpu and 1 GB ramI also have a test setup running on ESXi 5.1.0 build 799733, on this setup the problem is not there and the cpu is normal.
Almost the same vlan setup.I have been looking a the differences between the two but so far no luck.
-
… this is a cloud server so build is unknow.
Hope you're charged by CPU time.
It will be interesting to see what Deadringers is using but I have no idea where to look next.
-
pfsense is running 98% idle, ESXi is reporing 75% load
Then i run a iperf, pfsense reports 75% idle and ESXi 100% load getting 405mbit/sec
another test, two iperf sessions running to different vlan
pfsense going to 60% idle, so clearly getting more cpu, again esx is on 100%
iperf getting 412mbit/sec and 313Mbit/secSo ESXi is at 100% but this seems not to affect pfsense i terms of pushing more data
But at that time ESX is giving pfsense 3.6 Ghz -
Noebas and Deadringers,
What versions of pfSense and ESXi are you running?
How many vCPUs did you allocate to pfSense?
Hi there,
I am running esxi 5.1 and the latest stable PFsense build 2.0.3.
And I have 1 vCPU assigned to pfsense with 2 cores.
-
The hosting is running on ESXi, 5.0.0, 1117897
-
I have Pfsense 2.0.3 running om ESXi 5, build xxxx this is a cloud server so build is unknow.
1cpu and 1 GB ramI also have a test setup running on ESXi 5.1.0 build 799733, on this setup the problem is not there and the cpu is normal.
Almost the same vlan setup.I have been looking a the differences between the two but so far no luck.
Using the same version of ESXI as you are mate.
-
OK but my 5.1 test setup is doing fine. The 5.0 not.
-
This post seems to be the same issue : http://hardforum.com/showthread.php?t=1742250&page=2
I am willing to do any testing but this seems to be a bug in pfsense 2.x, as M0n0wall, also freebsd based, is not having this issue.
This is making pfsense 2.x unusable for us so i have to go back to 1.2.3 or some other solution. -
Can one of the admins comment on this or give me something to try?
I am downloading a file at about 70 mbps right now and my PFsense box is reporting about 10% CPU usage for "system"
However the idle usage is on 82% and ESXI is reporting my PFsense box is using up a huge amount of CPU.
-
I'm having this same issue. Is it safe to assume that nobody has found the root cause?
-
I'm having this same issue. Is it safe to assume that nobody has found the root cause?
In a word… No!
I have my suspicions that esxi is reporting high cpu because it is using a lot of cpu to just pass the traffic / packets but the router is fine as it sees it's nics as just nics and not a cpu.
The performance is fine for me right now with a 80 down 20 up connection.
-
I just did a test. On my pfsense VM have 1 dual core vCPU and 2 vNICs. One is for WAN and one is for LAN (both are setup as e1000 cards). Normally if I go onto the pfsense VM and do top -SH, I see 50% System CPU usage. {idle: cpu1} is consistent 65% {idle: cpu2} is consistent 35%.
I removed all vNICs and setup one of my physical PRO1000 intel adapters for PCI passthrough. I added this one PCI adapter to my pfsense VM. So it had NO vNICs at all. I configured pfsense for em0 as LAN and em_vlan50 as my WAN (WAN is on vlan 50).
End result: NO IMPROVEMENT
{idle: cpu1} is consistent 50% {idle: cpu2} is consistent 50%.
The system was still 50% utilized and uses 2600MHz on my esxi host. I do use pfsense as an OpenVPN client. not sure if that factors in. Still, no other processes on pfsense use hardly any cpu. Looks like all system overhead. Also, I'm on esxi 5.0 Update 2 because 5.1 broke all kinds of support for PCI passthrough. -
OK, I think my issue has been solved. I downloaded the ova for pfsense and did a vanilla install with no packages. I didn't see any CPU usage issues so I started comparing settings to my high CPU install and found one option that somehow got turned on at some point.
System–>Advanced
Enable device polling was checked onI disabled it and as soon as I did I started getting watchdog errors. I logged onto the pfsense console and rebooted the VM. After reboot, my CPU usage went from 2600MHz to a mere 138MHz. Hopefully some of you have the same option enabled. Not sure how it got turned on.