summary refs log tree commit diff stats
path: root/results/classifier/118/performance/1798057
diff options
context:
space:
mode:
Diffstat (limited to 'results/classifier/118/performance/1798057')
-rw-r--r--results/classifier/118/performance/179805772
1 files changed, 72 insertions, 0 deletions
diff --git a/results/classifier/118/performance/1798057 b/results/classifier/118/performance/1798057
new file mode 100644
index 00000000..0945dfd8
--- /dev/null
+++ b/results/classifier/118/performance/1798057
@@ -0,0 +1,72 @@
+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.
+