[SOLVED] PVE 4 CPU Cores

[sigxcpu]

What does /proc/meminfo show when you use transparent hugepages?

LE: Sorry, I've seen above. 8GB only. Is that with the VM on?

Start the VM (remove the mem-path thing and set vm.hugepages to 0) and post the whole contents of /proc/meminfo please.

 

[adamb]

What does /proc/meminfo show when you use transparent hugepages?

LE: Sorry, I've seen above. 8GB only. Is that with the VM on?

Start the VM (remove the mem-path thing and set vm.hugepages to 0) and post the whole contents of /proc/meminfo please.

Here it is. This is with the VM on. With the VM Off it shows as 0 for "AnonHugePages".

root@testprox1:~# cat /proc/meminfo | grep HugePages
AnonHugePages:    598016 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0

Here is the full output.

root@testprox1:~# cat /proc/meminfo 
MemTotal:       396142204 kB
MemFree:        383848600 kB
MemAvailable:   383822620 kB
Buffers:           35248 kB
Cached:           186516 kB
SwapCached:            0 kB
Active:          8411100 kB
Inactive:         110584 kB
Active(anon):    8312104 kB
Inactive(anon):    69420 kB
Active(file):      98996 kB
Inactive(file):    41164 kB
Unevictable:        8352 kB
Mlocked:            8352 kB
SwapTotal:      36569084 kB
SwapFree:       36569084 kB
Dirty:                96 kB
Writeback:             0 kB
AnonPages:       8312508 kB
Mapped:           112912 kB
Shmem:             77420 kB
Slab:             142904 kB
SReclaimable:      33120 kB
SUnreclaim:       109784 kB
KernelStack:       11680 kB
PageTables:        28640 kB
NFS_Unstable:          0 kB
Bounce:                0 kB
WritebackTmp:          0 kB
CommitLimit:    234640184 kB
Committed_AS:   340342784 kB
VmallocTotal:   34359738367 kB
VmallocUsed:     3506448 kB
VmallocChunk:   34154744748 kB
HardwareCorrupted:     0 kB
AnonHugePages:    661504 kB
CmaTotal:              0 kB
CmaFree:               0 kB
HugePages_Total:       0
HugePages_Free:        0
HugePages_Rsvd:        0
HugePages_Surp:        0
Hugepagesize:       2048 kB
DirectMap4k:      200700 kB
DirectMap2M:     5959680 kB
DirectMap1G:    398458880 kB

 

[adamb]

Doing some more testing. If I set "ACPI Support" to No it no longer goes through "kvm-clock: cpu 1" for each CPU and doesn't create a huge load. I still run into the high CPU issue as soon as the VM is powered on but I think this is simply reserving the large amount of memory. Any ideas whats going on with ACPI? Maybe I will try an older kernel.

 

[dietmar]

Maybe I will try an older kernel.

You already testes 4.2.3-1-pve from pvetest?

 

[adamb]

You already testes 4.2.3-1-pve from pvetest?

Trying the test kernel now.

On a side note. A CentOS7 VM with anything more than 150G of ram won't even boot, it actually throws an NMI on the VM. CentOS6 for whatever reason will actually boot most of the time.

 

[adamb]

Trying the test kernel now.

On a side note. A CentOS7 VM with anything more than 150G of ram won't even boot, it actually throws an NMI on the VM. CentOS6 for whatever reason will actually boot most of the time.

Latest kernel is no different. However I can confirm that the issue does not exist in the 3.9 or 4.1 kernel.

This sure does sound like our issue.

https://lists.gnu.org/archive/html/qemu-devel/2015-10/msg00100.html

 

[adamb]

Tested the following kernel with latest packages and the issue still persists. Go back to 4.1 or 3.9 and the issue is gone.

root@testprox2:~# uname -r
4.2.3-2-pve

 

[adamb]

Just wanted to update this thread. The following kernel fixed this issue.

pve-kernel-4.2.3-2-pve: 4.2.3-22