On 08/10/2010 01:06 AM, Thomas Fjellstrom wrote: > On August 9, 2010, Cole Robinson wrote: >> On 08/08/2010 02:57 AM, Thomas Fjellstrom wrote: >>> Hi, >>> >>> I've tried on several occasions to use the virt-manager to set memory >>> in a guest at run time, but all it seems to do is set memory in the >>> guest to a very small amount regardless of what number I actually >>> chose (trying to use a number in KB doesn't work, the gui seems to >>> have issues with it. it'll set the max mem, but trying to set mem to >>> anything that large will just make it go back to the previous >>> allocation). >>> >>> I just tried it again with virt-manager, memory was set to 256M, I told >>> it to increase to 512MB, and it instantly dropped to 70MB and the >>> guest OOMKilled every process. Then I forced the guest off, then tried >>> to restart it and virt-manager decided to say it was going to set the >>> memory to 1024MB, then crash. Though it did start the guest, but only >>> with the 512MB I increased it to in the last step. >>> >>> virsh seems to work fine to reduce the memory in a vm. I did notice >>> that virsh takes KB and virt-manager takes MB. Is it possible that >>> virt-manager isn't doing the necessary conversion to KB when calling >>> into libvirt? >> >> More info is really required here: versions for virt-manager, virtinst, >> libvirt, and qemu or xen depending on what you are using. If you can >> provide the output of virt-manager --debug when reproducing this series >> of events it would help track down the issue. > > libvirt-bin 0.8.2-1 > virt-manager 0.8.4-7 > virtinst 0.500.3-2 > > qemu 0.12.5 > qemu-kvm 0.12.4+ > linux 2.6.34.1 > > I've attached a log of the --debug output of virt-manager for a session > where I start it up, it connects to two of my servers (boris, kvm based, the > one that likes to not set memory properly, and snidely xen based that we'll > ignore for now) >
Hmm, seems like virt-manager is using the correct values, at least for redefining the persistent VM config. > After I kill the vm in that log you'll see I try to set the memory again, > and restart the vm, but this time it just didn't start. it said it was > running, but virt-manager didn't want to connect to the console and the > stats for it were stuck, all I could do was "force off" again. > Not sure what happened there, could be a weird side effect of the messed up memory values. Can you retry the memory change on a running VM, then provide: virsh dumpxml $vmname virsh dumpxml --inactive $vmname virsh dominfo $vmname What virsh command were you using for memory hotplug that worked successfully? Can you try that too starting from a working config, then provide the virsh output as above. > Actually I'm confused as to why the xen one isn't reading disk usage > properly. Maybe it doesn't like that the xen setup is LVM backed rather than > file backed? But then the kvm instance is lvm backed too and that works fine. > Not sure either. I'd file a xen bug with your distro. - Cole _______________________________________________ virt-tools-list mailing list [email protected] https://www.redhat.com/mailman/listinfo/virt-tools-list
