September 2012 - Users - libvirt List Archives

[libvirt-users] issue with kvm host crashing
by David M. Barlieb 11 Sep '12

11 Sep '12

I have a dell r910 with rhel6.2 on it and libvirt 0.8. This machine is hosting 12 virtual guests. Every 3 - 5 months the server crashes for no apparent reason. The logs show no kernel panics or other issues causing the crash. The sar logs show a very high context switch count ( approx. 170000). and also high runq-sz (approx. 10 - 18). The cpu's were mostly idle, memory usage low, no swapping, disk io very low as well. This also occurs on a number of other rhel6.2 servers I have using KVM/libvirt/qemu for virtualization. I am curious if anyone else has reported incidents like this. After the crash the servers all come backup, but as you can imagine, it is troubling to see this kind of behavior, especially with these machines hosting production guests. Any help or suggestions on what to look for would be helpful. Regards

1 0

[libvirt-users] dumpxml and edit the vm xml are not match.
by Timon Wang 11 Sep '12

11 Sep '12

I use vnc with password for the vm graphics, it works well. When I want to dump the xml of the vm, I found that the dumped xml does not contains the password properties in the graphics node. But when I use virsh edit to check the xml of the vm, I found the password is displayed. It's it a bug or it's just what libvirt designed to perform. -- Focus on: Server Vitualization, Network security,Scanner,NodeJS,JAVA,WWW Blog: http://www.nohouse.net

2 2

[libvirt-users] Reg: Libvirt API does not return, after qemu-kvm is hungup.
by Alphonse Hansel Anthony 10 Sep '12

10 Sep '12

Hi All, I am facing the following issue, while using Libvirt version 0.8.4. (1) Libvirt API does not return, since the qemu-kvm process is hung up. Based on the information available on mail thread *https://bugzilla.redhat.com/show_bug.cgi?id=676205* (comment 26 - scenario 1), this has been fixed in later version of Libvirt, specifically in RHEL 6.2(libvirt 0.9.4). I have checked the changelogs between version 0.8.7(RHEL 6.1) & 0.9.4 and was not able to identify the related patch information. Please let me know, If I am looking at the wrong version for this fix? Any help in this regard, would be useful. Thanks, Alphonse Comment reference: --SNIP-- 1. The QEMU process has hung. QEMU won't respond to monitor commands. The API call making the first monitor command will wait forever, any subsequent API calls issuing monitor commands will timeout after ~30 seconds with this libvirt error message. This is expected behaviour when QEMU has hung. --SNIP-- In RHEL-6.2 we have done a number of things to address / mitigate these problems - It is now always possible to destroy a guest, even if the monitor is stuck. This lets you destroy a guest in scenario 1, which is not always possible with RHEL-5 libvirt, without restarting libvirtd. --SNIP-- --EOF--

2 1

[libvirt-users] How to properly test watchdog?
by Russell Jones 10 Sep '12

10 Sep '12

CentOS 6 Hi all, I am working on setting up sanlock + watchdog on a 2 node KVM pair. Sanlock is working beautifully and is preventing access to the VM disks by more than one process as it should across both boxes. I am attempting to test failure scenarios involving watchdog, but I am having a hard time getting it to actually reset the server. I am running wdmd with -D so I can see the register requests. When I start sanlock with "service sanlock start" I can see it talk to the wdmd process, as debug logs "register ............ sanlock_daemon". I then send a "kill -9" signal to the daemon to try to simulate a crash, and wdmd logs "client_pid_dead". I would expect Watchdog to see that the PID is dead and as a result start the timer and reboot the server. It does not do that however. Am I misunderstanding how it reacts to TERM requests? How can I properly test wdmd? Thanks for the help!

1 1

[libvirt-users] Kernel unresponsive after booting 700+ vm's on a single host
by Alfred Bratterud 10 Sep '12

10 Sep '12

For a research project we are trying to boot a very large amount of tiny, custom built VM's on KVM/ubuntu. The maximum VM-count achieved was 1000, but with substantial slowness, and eventually kernel failure, while the cpu/memory loads were nowhere near limits. Where is the likely bottleneck? Any solutions, workarounds, hacks or dirty tricks? A few more details here (tumbleweed question), and the possibility of an upvote: http://stackoverflow.com/questions/12243129/kvm-qemu-maximum-vm-count-limit Any tips would be much appreciated! Best regards, Alfred Bratterud Assistant Professor Dept. of Computer Science, Oslo and Akershus University College of Applied Sciences P: (+47) 2245 3263 M: (+47) 4102 0222

2 1

[libvirt-users] libvirt failed to respond after quite a few vnet bridges were created
by Yih Chuang 09 Sep '12

09 Sep '12

During concurrent 80 VM creation, each VM takes two network connections, libvirtd failed to respond. Neither restart libvirtd nor restart CIMOM helped. There were 152 vnet bridges found on this RHEL KVM host. After all the vnet bridges were manually removed, libvirtd responded normally again. Is there a known issue about the number limit of vnet bridges regarding libvirtd? Thank you. --Yih

2 1

[libvirt-users] boot device
by vivek hari 07 Sep '12

07 Sep '12

Hi, I am trying to create a guest machine on xen virtualized environment and install rhel on it by booting from cdrom. I face a issue in booting device. I gave boot order as hd,cdrom. When doing maiden installation, it will find nothing in hard disk and boots from cd. Once installation is completed and when it boots next time it boots from hard disk. It is the expected behaviour. But it fails to boot from hard disk and again boots from cdrom. I also tried specifying per boot device option. I specified disk image as boot order 1 and cd image as boot order 2. But when i power on for the first time, it tries to boot from harddisk and finds nothing there and says " No bootable device. Going to power off in 30 seconds". But expected is, it should boot from cd after trying hard disk. Below is my domain configuration xml file i used to create guest domain. <domain type='xen'> <name>vm-2</name> <memory>1048576</memory> <vcpu>1</vcpu> <os> <type arch='x86_64'>hvm</type> <loader>/usr/lib/xen/boot/hvmloader</loader>  </os> <features> <acpi/> <apic/> <pae/> </features> <on_poweroff>destroy</on_poweroff> <on_reboot>restart</on_reboot> <on_crash>restart</on_crash> <devices> <emulator>/usr/lib/xen/bin/qemu-dm</emulator> <disk type='file' device='disk'> <source file='/var/lib/libvirt/images/vm-2.img'/> <target dev='xvda' bus='ide' /> <boot order='1'/> <address type='drive' controller='0' bus='0' target='0' unit='0'/> </disk> <disk type='file' device='cdrom'> <source file='/var/lib/libvirt/images/Rhel53.iso'/> <target dev='hdc' bus='ide' /> <boot order='2'/> <readonly/> <address type='drive' controller='0' bus='1' target='0' unit='0' /> </disk> <interface type='bridge'> <source bridge='br0'/> </interface> <graphics type='vnc' port='-1' autoport='yes'/> </devices> </domain> I don't know where i am going wrong. So need help in resolving this issue. Thanks,

2 4

[libvirt-users] Live Block Migration with additional attached storage
by Chun-Hung Chen 07 Sep '12

07 Sep '12

Dear all, I am planning to use live block migration with one VM running on local disk and also attached additional disk from iSCSI or other shared storage. When block migration, not only local VM disk is copied to the destination but also the attached additional disk from shared storage. It is not desired in this situation. I just want the local VM disk is copied. Is there anyway to do this scenario? Does the concept of storage pool help this? I browse the source code but don't find hints right now. Thanks. Regards, Arnose

2 5

[libvirt-users] QEMU cmdline for default -smp reasonable (qemuBuildSmpArgs) ?
by sanjay 07 Sep '12

07 Sep '12

Hi! When only the vCPU tag is specified, libvirt by default generates a multi-socket topology(sockets=vCPUs,cores=1,threads=1) for the qemu command line. Is there a specific reason behind choice of this default behavior? Could a default of (socket=1,cores=vCPUs,threads=1) would have been a better choice ? -- Regards, Sanjay

2 1

[libvirt-users] defining directly to the wan/lan
by gary artim 07 Sep '12

07 Sep '12

I have a fedora system configured and running with virsh using kvm/qemu. All I want is to have a fixed ipaddress for that virtual machine. I know that there is the Host and Guest sides to the network. The host has virbr0 interface, which i guess is a bridge. Seem when ever I define a interface to the guest with a type=routed it hoses my other interfaces on the host and requires a console reboot -- completely hanging the network. I'm no networking expert, but there must be a simple N step procedure for defining a static IPaddress to route to a guest machine? Has anyone got this working and have a procedure. My network has 2 nic, eth0 and eth1 (10.0.0.253 and 10.0.1.253), one nic has a route to an nfs machine, the other passes through a linux based router to the wan. I can alias the 10.0.1.253 (eth0) to another address in the subnet, like 10.0.1.251 (eth0:0) and would like to use it in some way to connect to the guest -- I have the nat/routing working fine from the router to eth0:0. The network for libvirt is confusing. It not obvious which side (guest or host) your defining and seem to keep multiple definitions around, even after a restart of libvirtd.service. An help would be great!

1 0