blob: 0945dfd812eb0091d89f0370a840038773ebc7e4 (
plain) (
blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
|
performance: 0.812
KVM: 0.665
architecture: 0.657
files: 0.650
hypervisor: 0.637
graphic: 0.612
semantic: 0.609
kernel: 0.573
debug: 0.565
device: 0.479
ppc: 0.473
user-level: 0.412
x86: 0.412
socket: 0.376
network: 0.329
risc-v: 0.326
mistranslation: 0.301
register: 0.296
PID: 0.287
vnc: 0.284
boot: 0.256
arm: 0.229
permissions: 0.216
VMM: 0.206
peripherals: 0.198
virtual: 0.179
TCG: 0.177
assembly: 0.092
i386: 0.081
Not able to start instances larger than 1 TB
Specs:
CPU: Intel(R) Xeon(R) Gold 6132 CPU @ 2.60GHz
OS: Ubuntu 18.04 AMD64
QEMU: 1:2.11+dfsg-1ubuntu7.6 (Ubuntu Bionic Package)
Openstack: Openstack Queens (Ubuntu Bionic Package)
Libvirt-daemon: 4.0.0-1ubuntu8.5
Seabios: 1.10.2-1ubuntu1
The Problem:
We are not able to start instances, which have a memory size over 1 TB.
After starting the instance, they shortly lock up. Starting guests with a lower amount of RAM works
perfectly. We dealt with the same problem in the past with an older Qemu Version (2.5) by patching some source files according to this patch:
https://git.centos.org/blob/rpms!!qemu-kvm.git/34b32196890e2c41b0aee042e600ba422f29db17/SOURCES!kvm-fix-guest-physical-bits-to-match-host-to-go-beyond-1.patch
I think we now have somewhat the same problem here, however the source base changed and I'am not able to find the corresponding snippet to patch this.
Also, guests show a wrong physical address size which is probably the cause of the lock ups on high memory guests:
root@debug:~# grep physical /proc/cpuinfo
physical id : 0
address sizes : 40 bits physical, 48 bits virtual
Any way to fix this?
Hi Alex,
You should be able to fix this by passing the right cpu flags, e.g.:
-cpu IvyBridge,host-phys-bits=yes
or
-cpu IvyBridge,physbits=46
Dave
I'm assuming that the right physbits setting fixed the bug? ... so I'm marking this ticket as "Invalid". If the problem still persists, then please open again.
|