Re: [one-users] KVM process at 99% CPU
I'm not aware of the differences of those boot loaders but most probably one of them does not use the "halt" instruction to make the CPU sleep. On Fri, Nov 8, 2013 at 10:30 AM, Daniel Dehennin wrote: > Javier Fontan writes: > >> Debian live is a linux, the boot menu is not and eats all CPU. > > [...] > >>> Something is strange with this ISO, a Debian Live does not behave the same. > > Sorry, that's not what I meant, the “debian live boot menu” does not > behave like the “Ubuntu boot menu”. > > Regards. > -- > Daniel Dehennin > Récupérer ma clef GPG: > gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF > > ___ > Users mailing list > Users@lists.opennebula.org > http://lists.opennebula.org/listinfo.cgi/users-opennebula.org > -- Javier Fontán Muiños Developer OpenNebula - The Open Source Toolkit for Data Center Virtualization www.OpenNebula.org | @OpenNebula | github.com/jfontan ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] KVM process at 99% CPU
Javier Fontan writes: > Debian live is a linux, the boot menu is not and eats all CPU. [...] >> Something is strange with this ISO, a Debian Live does not behave the same. Sorry, that's not what I meant, the “debian live boot menu” does not behave like the “Ubuntu boot menu”. Regards. -- Daniel Dehennin Récupérer ma clef GPG: gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF pgpq6VgggkGCR.pgp Description: PGP signature ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] KVM process at 99% CPU
Debian live is a linux, the boot menu is not and eats all CPU. On Thu, Nov 7, 2013 at 9:56 PM, Daniel Dehennin wrote: > Javier Fontan writes: > >> Most, if not all, modern operating systems set the CPU to sleep mode >> when they don't need to be used. The problem here is that the VMs are >> not running a OS but a boot loader and it does not implement this >> feature so it is using all the CPU cycles doing nothing. In this >> regard KVM nor OpenNebula can help. > > Something is strange with this ISO, a Debian Live does not behave the same. > >> The only tip I can give you is to use cgroups to limit the total >> amount of CPU a VM can use: >> >> http://opennebula.org/documentation:rel4.2:kvmg#working_with_cgroups_optional > > I'll do that, it's not serious if users can overload a node with a fork > bomb in a VM ;-) > > Regards. > -- > Daniel Dehennin > Récupérer ma clef GPG: > gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF > > ___ > Users mailing list > Users@lists.opennebula.org > http://lists.opennebula.org/listinfo.cgi/users-opennebula.org > -- Javier Fontán Muiños Developer OpenNebula - The Open Source Toolkit for Data Center Virtualization www.OpenNebula.org | @OpenNebula | github.com/jfontan ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] KVM process at 99% CPU
Javier Fontan writes: > Most, if not all, modern operating systems set the CPU to sleep mode > when they don't need to be used. The problem here is that the VMs are > not running a OS but a boot loader and it does not implement this > feature so it is using all the CPU cycles doing nothing. In this > regard KVM nor OpenNebula can help. Something is strange with this ISO, a Debian Live does not behave the same. > The only tip I can give you is to use cgroups to limit the total > amount of CPU a VM can use: > > http://opennebula.org/documentation:rel4.2:kvmg#working_with_cgroups_optional I'll do that, it's not serious if users can overload a node with a fork bomb in a VM ;-) Regards. -- Daniel Dehennin Récupérer ma clef GPG: gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF pgpDbQx7i4Iq_.pgp Description: PGP signature ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
Re: [one-users] KVM process at 99% CPU
Most, if not all, modern operating systems set the CPU to sleep mode when they don't need to be used. The problem here is that the VMs are not running a OS but a boot loader and it does not implement this feature so it is using all the CPU cycles doing nothing. In this regard KVM nor OpenNebula can help. The only tip I can give you is to use cgroups to limit the total amount of CPU a VM can use: http://opennebula.org/documentation:rel4.2:kvmg#working_with_cgroups_optional On Thu, Nov 7, 2013 at 4:38 PM, Daniel Dehennin wrote: > Hello, > > On our 3.8 dual-node ONE I got this morning a OVERLOADED node with load > average of 35. > > I found one my own VMs eating 99% CPU, it was just booted and waiting > at ISO boot menu prompt. > > It looks to be related to KVM as I reproduce it on a 4.2. > > I can reproduce it by defining a template using an Ubuntu Server AMD64 ISO > and instanciate it. > > I just let the VMs at ISO boot menu, and the KVM process quickly eat 99% > of CPU. > > If someone has an idea. > > Here is the template: > > oneadmin@nebula:~$ onetemplate show 17 > TEMPLATE 17 INFORMATION > ID : 17 > NAME : Install ISO AMD64 20G > USER : eole > GROUP : oneadmin > REGISTER TIME : 10/11 19:29:17 > > PERMISSIONS > OWNER : um- > GROUP : --- > OTHER : --- > > TEMPLATE CONTENTS > CONTEXT=[ > NETWORK="YES", > SSH_PUBLIC_KEY="$USER[SSH_PUBLIC_KEY]" ] > CPU="0.2" > DISK=[ > IMAGE_ID="0" ] > DISK=[ > CACHE="none", > IMAGE_ID="43" ] > FEATURES=[ > ACPI="yes", > PAE="no" ] > GRAPHICS=[ > KEYMAP="fr", > LISTEN="0.0.0.0", > TYPE="VNC" ] > INPUT=[ > BUS="usb", > TYPE="tablet" ] > MEMORY="768" > NIC=[ > MODEL="virtio", > NETWORK_ID="1" ] > OS=[ > ARCH="x86_64", > BOOT="hd" ] > VCPU="2" > > And here the ISO: > > oneadmin@nebula:~$ oneimage show 43 > IMAGE 43 INFORMATION > ID : 43 > NAME : Ubuntu 13.10 server AMD64 > USER : eole > GROUP : oneadmin > DATASTORE : ISO > TYPE : CDROM > REGISTER TIME : 11/07 16:29:06 > PERSISTENT : No > SOURCE : > /var/lib/one//datastores/100/39670c51a99879ca3e221d5fc8027dce > PATH : /var/tmp/thin-body20131107-1782-yqfomp > SIZE : 672M > STATE : used > RUNNING_VMS: 1 > > PERMISSIONS > OWNER : um- > GROUP : --- > OTHER : --- > > IMAGE TEMPLATE > DEV_PREFIX="hd" > DRIVER="raw" > > VIRTUAL MACHINES > > ID USER GROUPNAMESTAT UCPUUMEM HOSTTIME >117 eole oneadmin userver-testrunn 99768M nebula 0d > 00h06 > > Regards. > > -- > Daniel Dehennin > Récupérer ma clef GPG: > gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF > > ___ > Users mailing list > Users@lists.opennebula.org > http://lists.opennebula.org/listinfo.cgi/users-opennebula.org > -- Javier Fontán Muiños Developer OpenNebula - The Open Source Toolkit for Data Center Virtualization www.OpenNebula.org | @OpenNebula | github.com/jfontan ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org
[one-users] KVM process at 99% CPU
Hello, On our 3.8 dual-node ONE I got this morning a OVERLOADED node with load average of 35. I found one my own VMs eating 99% CPU, it was just booted and waiting at ISO boot menu prompt. It looks to be related to KVM as I reproduce it on a 4.2. I can reproduce it by defining a template using an Ubuntu Server AMD64 ISO and instanciate it. I just let the VMs at ISO boot menu, and the KVM process quickly eat 99% of CPU. If someone has an idea. Here is the template: oneadmin@nebula:~$ onetemplate show 17 TEMPLATE 17 INFORMATION ID : 17 NAME : Install ISO AMD64 20G USER : eole GROUP : oneadmin REGISTER TIME : 10/11 19:29:17 PERMISSIONS OWNER : um- GROUP : --- OTHER : --- TEMPLATE CONTENTS CONTEXT=[ NETWORK="YES", SSH_PUBLIC_KEY="$USER[SSH_PUBLIC_KEY]" ] CPU="0.2" DISK=[ IMAGE_ID="0" ] DISK=[ CACHE="none", IMAGE_ID="43" ] FEATURES=[ ACPI="yes", PAE="no" ] GRAPHICS=[ KEYMAP="fr", LISTEN="0.0.0.0", TYPE="VNC" ] INPUT=[ BUS="usb", TYPE="tablet" ] MEMORY="768" NIC=[ MODEL="virtio", NETWORK_ID="1" ] OS=[ ARCH="x86_64", BOOT="hd" ] VCPU="2" And here the ISO: oneadmin@nebula:~$ oneimage show 43 IMAGE 43 INFORMATION ID : 43 NAME : Ubuntu 13.10 server AMD64 USER : eole GROUP : oneadmin DATASTORE : ISO TYPE : CDROM REGISTER TIME : 11/07 16:29:06 PERSISTENT : No SOURCE : /var/lib/one//datastores/100/39670c51a99879ca3e221d5fc8027dce PATH : /var/tmp/thin-body20131107-1782-yqfomp SIZE : 672M STATE : used RUNNING_VMS: 1 PERMISSIONS OWNER : um- GROUP : --- OTHER : --- IMAGE TEMPLATE DEV_PREFIX="hd" DRIVER="raw" VIRTUAL MACHINES ID USER GROUPNAMESTAT UCPUUMEM HOSTTIME 117 eole oneadmin userver-testrunn 99768M nebula 0d 00h06 Regards. -- Daniel Dehennin Récupérer ma clef GPG: gpg --keyserver pgp.mit.edu --recv-keys 0x7A6FE2DF pgp2fs2gWG7nw.pgp Description: PGP signature ___ Users mailing list Users@lists.opennebula.org http://lists.opennebula.org/listinfo.cgi/users-opennebula.org