- Devel - libvirt List Archives

[libvirt] feature or bug?
by Gene Czarcinski 30 Jun '09

30 Jun '09

Yet another question. I am using the virt-manager GUI. When I begin to define a new guest and attempt to "browse" to specify where the iso image file is, the GUI dialog brings me up in my home directory. But, almost everything associated with libvirt is done as the root user. By default, why isn't the initial directory "/var/lib/libvirt/images"? Gene

2 1

[libvirt] your patence is appreciated
by Gene Czarcinski 30 Jun '09

30 Jun '09

I really hate being a bit of a PITA with all of my question/comments, but I am just trying to use Fedora virtualization in a manner where I can reasonably manage the configuration. I suspect I am not the first ex-VMware user and I will not be the last such user who is trying to use Fedora virtualization in place of VMware. I am very familiar with how guests are "managed" with VMware. I want Fedora (and Linux) virtualization to be a good product. Your patience is appreciated! Gene

2 1

[libvirt] need some help/asdive
by Gene Czarcinski 30 Jun '09

30 Jun '09

Background: I am a newbie to Linux Virtualization. However, I have been using Linux for a long time. After a short time with Slackware, I started with Red Hat Linux 2 and went on to 2.1, 3.0.3, 4, 4.2, 5, 5.2, ..., 8, 9. And then on to Fedora Core 1, 2, ... until I installed Fedora 11 about a month ago. I have also run VMware starting when it first appeared. Initially I used VMware to run Windows. These days I use it to run up to from 25 to 30 guests (between 1 and 5 at the same time) [yes, I have lots of system memory]. Besides a couple of Windows guests, I have some guests supporting both 32 bit (i386) and 64 bit (x86_64) architectures of older releases as well as alternatives such as CentOS -- I use these to build RPMs. Then there are the guests supporting other operating systems such as Solaris and *BSD. Finally, I have a set of test guest that I use for testing stuff which could destroy the operating system or which require a lot of re-booting. All of this in 25 to 30 guests. With Fedora 11, I started by trying VMware-server. Server 1.0.9 (which I used on Fedora 9) looks like much more work to get running ... more or less "a bridge too far". So, I tried server 2.0.2 ... but the client interface changed from a separate GUI application to a browser based interface: 1) this impacts my usage of the firefox browser; 2) it is a buggy implementation; and 3) it is a crappy interface which I do not like. So, I (finally) noticed that Linux in general and Fedora specifically has packages supporting running guests virtually ... so, lets change my "problem sets" and give this a try ... especially when I noticed that it could use the hardware support in my new AMD Phenom II 940 processor. Before getting into my specific questions/problem, let me describe how my systems are set up. First of all, I have a basic rule: "don't screw with a working system." Therefore, on each of my systems I have at least two root partitions defined -- one for the current (working) system and one for the next (new) system [these days it is actually a /boot partition with root being on a striped logical volume]. I also have a small (think minimal install) partition with Fedora to serve as a boot selector (using chainloader), as something to to define striped logical volumes, and to be there for "emergency" maintenance. Guys, this works for me and always doing fresh installs saves on other headaches. Implicit in the above is that I put all of my data and my /home into separate partitions or logical volumes. One of these are my VMware guest systems. Now VMware puts all of the files (configuration, disk image, etc.) associated a guest under a separate directory. Fedora virtualization does not do this and separation is something I need so that I can easily move to a new version of the operating system such as when Fedora 12 becomes available in November. Questions/problems: My first question: it appears to me that a guest is defined by two things -- a configuration file in /etc/libvirt/qemu and one or more disk images in /var/lib/libvirt/images ... is this it or are there other files I need to consider? My second question concerns the location of the disk images: what is the "best practical way" to move these out of the root partition? I have come up with these different approaches which could/may work: 1. change the "path" value in /etc/libvirt/storage/default.xml to point to a directory in a different partition. 2. Use a /etc/fstab entry to "bind mount" a directory in a separate partition onto /var/lib/libvirt/images 3. Just use an /etc/fstab entry to mount a partition on /var/lib/libvirt/images or, perhaps, on /var/lib/libvirt. [This could be done during system installation] OK, any comments on these?? I can live with changing some configuration files, I can live with copying guest configuration files between systems, I CANNOT live with the disk image files being put into the root partition. Before jumping "whole hog" into using Fedora Virtualization in place of VMware, I need to be able to live with the result. Figure 30 guest virtuals with 10GB of virtual disk each, that is 300GB. I have that space but not for every root partition on my system! Comment: It sure would be nice if Fedora Virtualization put all files related to a guest under a single directory the way VMware does. Comment: It sure would be nice if there was some system-wide configuration parameter that defined where virtualization files went. Right now, as a newbie, I am using the virt-manage GUI interface. I am studying the command-line programs to see what they offer me. So far, I like what I see in Fedora Virtualization but I need to address how things will be practically managed sooner rather than later. Any help appreciated. Gene

2 4

[libvirt] PATCH: Remove unused Open Nebula code, fix potential missing NULL terminator
by Daniel P. Berrange 30 Jun '09

30 Jun '09

This patch removes a bunch of unused Open Nebula code which has a number of flaws. It also fixes handling of a strncpy() call to ensure there is a NULL terminator in the copied string. And finally it makes a couple of strings be 'const' since they're never changed Daniel Index: src/opennebula/one_client.c =================================================================== RCS file: /data/cvs/libvirt/src/opennebula/one_client.c,v retrieving revision 1.1 diff -u -p -r1.1 one_client.c --- src/opennebula/one_client.c 24 Jun 2009 17:32:56 -0000 1.1 +++ src/opennebula/one_client.c 30 Jun 2009 11:05:12 -0000 @@ -33,16 +33,9 @@ void c_oneStart() xmlrpc_client_init2(&one_client.env, XMLRPC_CLIENT_NO_FLAGS, "OpenNebula API Client", "1.2", NULL, 0); - one_client.error=0; - -#ifdef ONED_PORT - one_client.url=(char *)malloc(64); - snprintf(one_client.url, 63, "http://localhost:%d/RPC2", ONED_PORT); -#else - one_client.url=(char *)"http://localhost:2633/RPC2"; -#endif - - one_client.session=(char *)"one-session"; + one_client.error = 0; + one_client.url = "http://localhost:2633/RPC2"; + one_client.session = "one-session"; }; @@ -94,35 +87,6 @@ int c_oneMigrate(int vmid, int hid, int return c_oneReturnCode(resultP); } -int c_oneAllocate(char* template_file) -{ - int file; - int size; - int bytes_read; - char *file_text; - int return_code; - - file=open(template_file, O_RDONLY); - size=lseek(file, 0, SEEK_END); - lseek(file, 0, SEEK_SET); - - file_text=(char *)malloc(size+1); - bytes_read=read(file, file_text, size); - close(file); - - if(bytes_read==size) { - file_text[size]=0; - - return_code=c_oneAllocateTemplate(file_text); - free(file_text); - - return return_code; - } else { - free(file_text); - return -1; - } -} - int c_oneAllocateTemplate(char* vm_template) { xmlrpc_value *resultP; @@ -222,6 +186,7 @@ int c_oneVmInfo(int vmid, char* ret_info if( return_code ) { strncpy(ret_info, return_string, length-1); + ret_info[length-1] = '\0'; xmlrpc_DECREF(resultP); free(return_string); @@ -241,9 +206,6 @@ void c_oneFree() { xmlrpc_env_clean(&one_client.env); xmlrpc_client_cleanup(); -#ifdef ONE_PORT - free(one_client.url); -#endif } Index: src/opennebula/one_client.h =================================================================== RCS file: /data/cvs/libvirt/src/opennebula/one_client.h,v retrieving revision 1.1 diff -u -p -r1.1 one_client.h --- src/opennebula/one_client.h 24 Jun 2009 17:32:56 -0000 1.1 +++ src/opennebula/one_client.h 30 Jun 2009 11:05:12 -0000 @@ -24,8 +24,8 @@ struct _oneClient { xmlrpc_env env; - char *url; - char *session; + const char *url; + const char *session; char *error; }; @@ -38,8 +38,6 @@ int c_oneDeploy(int vmid, int hid); int c_oneMigrate(int vmid, int hid, int flag); -int c_oneAllocate(char* template_file); - int c_oneAllocateTemplate(char* vm_template); int c_oneAction(int vmid,char* action); -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

2 1

[libvirt] problem: deleting a guest
by Gene Czarcinski 30 Jun '09

30 Jun '09

I defined a new guest using the virt-manager GUI ... everything fine so far. I then decided that my definition was not correct (or I could be done with that guest) and could not find a way to delete the guest. OK, quit virt-manager, delete the configuration and disk-image files for the guest, restart libvirtd, and then start virt-manager again. The guest definition (according to virt-manager) was still there. I finnaly got rid of the definition by re-booting the system. THIS IS NOT SATISFACTORY! IMHO, I should be able to: 1. Define a new guest. 2. Delete the definition of a new guest but not the related configuration and disk-image files. 3. Delete the guest definition and the related configuration and disk image files but NOT and CD/DVD image files external to /var/lib/libvirt/images. 4. Import new guests which were created as clones, copied from another host, or convert created from something like VMware. Not it is certainly possible (or even likely) that I h\just do not understand how to do things yet. Or, this could be a bug. Or, this could be a lack of capabilities. Help appreciated in pointing me in the right direction. I will bugzilla where appropriate. Gene

3 3

[libvirt] How does libvirt 0.6.3 load kvm driver
by Jason Wong 30 Jun '09

30 Jun '09

Dear all, I found that libvirt 0.6.3 seems does not load KVM driver while libvirt 0.3.3 does. Situation description: I am using CentOS 5.3, and the kernel I used is 2.6.30. I downloaded the kernel from kernel.org andcompile it. KVM was compiled as a module. libvirt: libvirt 0.6.3, released by Eucalyptus(a cloud computing infrastructure software which supports KVM as well) (1) libvirt 0.6.3 does not load KVM driver After booting the machine with kernel 2.6.30 and probe KVM related module. I started libvirtd 0.6.3 and set LD_LIBRARY_PATH to "$LIBVIRT_063_HOME/lib:$LIBVIRT_063_HOME/usr/lib:$LD_LIBRARY_PATH". I found that I am not able to connect to libvirtd using "virsh --connect qemu:///system"(the virsh used is virsh 0.6.3 that shipped with libvirt 0.6.3), here is the debug message: ======================================= debug message ========================================= 17:41:11.219: debug : virInitialize:290 : register drivers 17:41:11.219: debug : virRegisterDriver:667 : registering Test as driver 0 17:41:11.219: debug : virRegisterNetworkDriver:567 : registering Test as network driver 0 17:41:11.219: debug : virRegisterStorageDriver:598 : registering Test as storage driver 0 17:41:11.219: debug : virRegisterDeviceMonitor:629 : registering Test as device driver 0 17:41:11.219: debug : virRegisterDriver:667 : registering Xen as driver 1 17:41:11.219: debug : vboxRegister:68 : VBoxCGlueInit failed: 0.0.0, errorval=-1 17:41:11.219: debug : virRegisterDriver:667 : registering VBOX as driver 2 17:41:11.219: debug : virConnectOpen:1050 : name=qemu:///system 17:41:11.219: debug : do_open:920 : name "qemu:///system" to URI components: scheme qemu opaque (null) authority (null) server (null) user (null) port 0 path /system 17:41:11.219: debug : do_open:930 : trying driver 0 (Test) ... 17:41:11.219: debug : do_open:936 : driver 0 Test returned DECLINED 17:41:11.219: debug : do_open:930 : trying driver 1 (Xen) ... 17:41:11.219: debug : do_open:936 : driver 1 Xen returned DECLINED 17:41:11.219: debug : do_open:930 : trying driver 2 (VBOX) ... 17:41:11.219: debug : do_open:936 : driver 2 VBOX returned DECLINED libvir: error : could not connect to qemu:///system 17:41:11.219: debug : virUnrefConnect:210 : unref connection 0x60dd00 1 17:41:11.219: debug : virReleaseConnect:171 : release connection 0x60dd00 libvirt error: could not connect to qemu:///system (code=5) error: failed to connect to hypervisor ================================end of debug message ========================================= It seems that libvirt 0.6.3 does not load QEMU driver. (2) libvirt 0.3.3 does load KVM driver However, when I run the command "lsmod | grep kvm". I found KVM module was present: ================================= result of lsmod ============================================== [root@node2 ~]# lsmod | grep kvm kvm_intel 46664 0 kvm 166224 1 kvm_intel [root@node2 ~]# ================================= end of result of lsmod ========================================= I stopped libvirtd 0.6.3 and then started libvirtd 0.3.3(the one shipped with CentOS 5.3). The virtual machine manager works well and I can use "virsh --connect qemu:///system"(the virsh used is virsh 0.3.3 that shipped with CentOS 5.3) to connect to libvirtd. I dig into libvirt source code. I found that in libvirt 0.3.3: There is qemudRegister() in libvirt.c. I think this function call is responsible for loading qemu(kvm) driver. And in libvirt 0.6.3: I cannot found qemudRegister() or virDriverLoadModule("qemu") in the source code. My question is: Does libvirt 0.6.3 support KVM? If yes, how does it load KVM driver? I want to find out why libvirt 0.3.3 works well while libvirt 0.6.3 does not work in my computer. Thanks. -- Jason Wong System engineer Cluster Technology Limited Email: hcwong(a)clustertech.com Direct Line: (852) 2655-6129 Tel: (852) 2655-6100 Fax: (852) 2994-2101 Website: www.clustertech.com ------------------------------ ************************************************************************** The information and attachment contained in this e-mail originating from Cluster Technology Limited is confidential and intended solely for the specified addressees. If you have received this email in error, please do not read, copy, distribute, disclose or use any information of this email in any way and please immediately notify the sender and delete this email. Thank you for your cooperation. **************************************************************************

2 4

[libvirt] hooking a physical nic to a bridge with net-create
by Shahar Klein 29 Jun '09

29 Jun '09

Hi I trying to set up a network configuration using libvirt The setup is: 1. a VM using virtio using a tap device connected to a bridge device 2. on the host side eth0 added to the same bridge device(without ip) using libvirt application (say virsh) I want to launch this set up: virsh net-create for the second part virsh create for the first part(tap, virtio, vm etc) I can set the first part(vm, tap, virtio) using the domain xml but I'm also interested in XMLing the second part and net-create it. I tried but only suceeded in creating the bridge Is it possible? am I missing somthing? Thanks in advance

4 7

[libvirt] HACKING file on website
by Daniel P. Berrange 29 Jun '09

29 Jun '09

FYI, I just committed a patch to add the contents of the 'HACKING' file to the website, under the section 'Documentation -> Internals -> Contributor Guidelines No content changes from HACKING, just html-ized existing content. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

1 0

[libvirt] identi.ca micro-blogging group for libvirt & related apps
by Daniel P. Berrange 29 Jun '09

29 Jun '09

Micro-blogging is very popular these days and I know a number of people working on / with libvirt are using twitter/identi.ca So I decided it was worth creating a libvirt group on identi.ca for targetting notices relating to libvirt and any apps built using libvirt. You can find it here http://identi.ca/group/libvirt Any posting you make containing the magic word '!libvirt' will appear in the group page. Regards, Daniel -- |: Red Hat, Engineering, London -o- http://people.redhat.com/berrange/ :| |: http://libvirt.org -o- http://virt-manager.org -o- http://ovirt.org :| |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :| |: GnuPG: 7D3B9505 -o- F3C9 553F A1DA 4AC2 5648 23C1 B3DF F742 7D3B 9505 :|

1 0

[libvirt] [PATCH] Fixing issues in re-detection of transient VMs
by Federico Simoncelli 29 Jun '09

29 Jun '09

Re-detection of transient VMs had two issues described in bugs 507304 and 507537: https://bugzilla.redhat.com/show_bug.cgi?id=507304 Summary: After a migration the domain status is saved as "paused" and it is not updated to "running". A following libvirtd restart will detect the vm as "paused" (instead of "running"). https://bugzilla.redhat.com/show_bug.cgi?id=507537 Summary: Destroying a transient re-detected domain leaves the domain defined (instead of completely remove it). Patches in attachment (and available in the bug report). -- Federico Simoncelli.

2 2