On 10 Dec 2015, at 11:30, Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:
On Thu, Dec 10, 2015 at 10:41 AM, Michal Skrivanek <mskrivan@redhat.com> wrote:
On 10 Dec 2015, at 00:11, Gianluca Cecchi <gianluca.cecchi@gmail.com> wrote:
On Wed, Dec 9, 2015 at 10:21 PM, Sandro Bonazzola wrote:
No idea. If it's libvirt issue, can you test http://cbs.centos.org/koji/buildinfo?buildID=4726 ? it's from centos virt sig, introduced for Xen but should work on kvm as well. or the 7.2 libvirt from http://mirror.centos.org/centos/7/cr/x86_64/Packages/.
Then I reboot my HV, exit from maintenance, connect to engine and start VM. VM situation $ free total used free shared buff/cache available Mem: 8172900 147356 7875692 8480 149852 7847248 Swap: 839676 0 839676
I set memory from 8192Mb to 10240Mb I get the same window I got with previous libvirt version. I select OK (I don't select "Apply later" checkbox)
Inside guest there is no changed memory, but I can see this in /var/log/messages:
could it be perhaps the guest has troubles recognizing hotplug? Is it also a CentOS 7.1 guest?
Yes, CentOS 7.1 updated up to a couple of days ago.
Can you try 7.2? I don’t remember exactly but it may be that in earlier guests it’s not automatic. Check some tips for onlining memory explicitly 5.2. How to online memory ------------ Even if the memory is hot-added, it is not at ready-to-use state. For using newly added memory, you have to "online" the memory block. For onlining, you have to write "online" to the memory block's state file as: % echo online > /sys/devices/system/memory/memoryXXX/state
Dec 9 23:56:18 racclient1 kernel: init_memory_mapping: [mem 0x240000000-0x2bfffffff]
ANd output of dmesg contains:
[ 219.363640] ACPI: \_SB_.MP00: ACPI_NOTIFY_DEVICE_CHECK event [ 219.364248] init_memory_mapping: [mem 0x240000000-0x2bfffffff] [ 219.364253] [mem 0x240000000-0x2bfffffff] page 2M [ 219.366773] [ffffea0009000000-ffffea00091fffff] PMD -> [ffff8800b7e00000-ffff8800b7ffffff] on node 0 [ 219.368433] [ffffea0009200000-ffffea00093fffff] PMD -> [ffff880231400000-ffff8802315fffff] on node 0 [ 219.369838] [ffffea0009400000-ffffea00095fffff] PMD -> [ffff880233600000-ffff8802337fffff] on node 0 [ 219.371226] [ffffea0009600000-ffffea00097fffff] PMD -> [ffff880233000000-ffff8802331fffff] on node 0 [ 219.372790] [ffffea0009800000-ffffea00099fffff] PMD -> [ffff880232c00000-ffff880232dfffff] on node 0 [ 219.374185] [ffffea0009a00000-ffffea0009bfffff] PMD -> [ffff880232200000-ffff8802323fffff] on node 0 [ 219.377349] [ffffea0009c00000-ffffea0009ffffff] PMD -> [ffff8800b7400000-ffff8800b77fffff] on node 0 [ 219.378716] [ffffea000a000000-ffffea000a1fffff] PMD -> [ffff8800b7000000-ffff8800b71fffff] on node 0 [ 219.380115] [ffffea000a200000-ffffea000a3fffff] PMD -> [ffff880230c00000-ffff880230dfffff] on node 0 [ 219.388147] [ffffea000a400000-ffffea000abfffff] PMD -> [ffff880227c00000-ffff8802283fffff] on node 0 [ 219.389687] [ffffea000ac00000-ffffea000adfffff] PMD -> [ffff8800b7200000-ffff8800b73fffff] on node 0
but “free” sstill shows the same value as before hotplug?
Yes, output of "free" command show the same amount of memory
I then shutdown the VM and power on it again and I get the changed memory:
well, yeah, but that doesn’t count since you’ve shut it down, so the next run is initialized with 10GB
It was only to verify that after shutdown/power on the changed setting would have been applied to the VM
$ free total used free shared buff/cache available Mem: 10237276 167872 9919656 8480 149748 9891280 Swap: 839676 0 839676
BTW: When I press ok in the gui for memory increase I get these events in webadmin: Dec 9, 2015 11:56:22 PM VM racclient1 c71_Disk1_newtemplate disk was updated by admin@internal.
hm..doesn’t sound right. Did the confirmation window show any more fields as changed other than memory?
The window I got when I increased the amount of memory was the same as in a previous test where I selected "Cancel", before libvirt update; see https://drive.google.com/file/d/0BwoPbcrMv8mvbUw2VnpCMTNNQlU/view?usp=sharin...
And I also commented about misleading information inside it: I didn't clearly understand what would have been changed and what not, because I see memory information in both parts…..
I agree it’s confusing. Please do open a bug on that Thanks, michal
It doesn't seem as expected, does it?
I think we’re almost there. Just need to figure out what happened in the guest. I would suspect a problem there
Thanks, michal
Gianluca
The gui events have the same timestamp of the moment when I pressed ok. After 2-3 minutes I did shutdown and power on of the VM and I only got the down/up events... I'm available to make further test you need.
GIanluca