March 2009 - Devel - libvirt List Archives

[libvirt] Restoring from a largish 'virsh save' file
by Matt McCowan 26 Mar '09

26 Mar '09

Running into an issue where, if I/O is hampered by load for example, reading a largish state file (created by 'virsh save') is not allowed to complete. qemudStartVMDaemon in src/qemu_driver.c has a loop that waits 10 seconds for the VM to be brought up. An strace against libvirt when doing a 'virsh restore' against a largish state file shows the VM being sent a kill when it's still happily reading from the file. inotify is used in the xen and uml drivers, so I thought it would be a suitable mechanism to delay the timeout loop if the state file was still being read. Is this the right way to solve this problem? I've inlined some (flimsy) mods for comment Regards Matt McCowan --- libvirt.orig/src/qemu_driver.c 2009-03-20 11:17:57.000000000 +0900 +++ libvirt-0.6.1.1/src/qemu_driver.c 2009-03-22 18:36:06.000000000 +0900 @@ -43,6 +43,7 @@ #include <stdio.h> #include <sys/wait.h> #include <sys/ioctl.h> +#include <sys/inotify.h> #if HAVE_NUMACTL #define NUMA_VERSION1_COMPATIBILITY 1 @@ -1310,6 +1309,11 @@ pid_t child; int pos = -1; char ebuf[1024]; + char rpathname[PATH_MAX]; + int fd = 0; + int wd = 0; + int got = 0; + char buf[1024]; struct gemudHookData hookData; hookData.conn = conn; @@ -1430,6 +1434,19 @@ qemudSecurityHook, &hookData); /* wait for qemu process to to show up */ + + /* Times out if using a restore file and that file + is big. Use inotify */ + if (stdin_fd > 0) { + fd = inotify_init(); + + if (snprintf(rpathname, sizeof(rpathname), "/proc/self/fd/%d", + stdin_fd) >= (int)sizeof(rpathname)) { + return -1; + } + wd = inotify_add_watch(fd, rpathname, IN_ACCESS); + } + if (ret == 0) { int retries = 100; int childstat; @@ -1442,7 +1459,15 @@ if ((ret = virFileReadPid(driver->stateDir, vm->def->name, &vm->pid)) == 0) break; usleep(100*1000); - retries--; + + /* Is state file being accessed? */ + got = read(fd, buf, sizeof(buf)); + if (got < 1) + retries--; + } + if (stdin_fd > 0) { + inotify_rm_watch(fd, wd); + close(fd); } if (ret) { qemudReportError(conn, NULL, NULL, VIR_ERR_INTERNAL_ERROR,

2 3

Re: [libvirt] Using libvirt from PHP
by Russell Haering 25 Mar '09

25 Mar '09

* Sorry, I somehow sent this to the wrong address * Radek, Not to advertise blatantly, but how attached are you to PHP? I'm currently working on a Django (python) WebApp that (although its *very* pre-release) could probably be adapted to what you're doing. If you are interested in assisting in development you'd be welcome to, or if you just wanted to fork off the code into your own thing you could do that too. My system consists of a python daemon used for actual libvirt interaction and a separate django web interface that interacts with the daemon via AMF over https. If you want to stick with what you have now, you could try doing something similar with a python daemon (you really need something like that for the sake of efficiency anyway, unless you can come up with a better way of maintaining libvirt connections from a stateless webapp) that your PHP frontend connects to. Or you could just steal my daemon code, although at the moment all it really does is monitoring and (as of yesterday) live migrations. What I have so far: http://trac.osuosl.org/trac/virtadmin -- Russell Haering Systems Administrator OSU Open Source Lab Radek Hladik wrote: > > Hi, > > I am developing wimple web application in PHP to monitor and control > > VMs using libvirt. I hope it could become simple library for doing basic > > things with VMs. And I would like to ask for your opinion about > > different possibilities of calling libvirt. > > > > Option 1) I am calling virsh via exec command now (one virsh for every > > libvirt command). This can be used for local libvirt but it is very > > inefficient. The overhead for remote libvirt would be unbearable. I > > could run one virsh per page and use it "interactively" so the overhead > > gets lower. > > Option 2) The other option is to create Zend extension which should be > > able to call libvirt directly from C. However I've never done anything > > like that so I do not know, whether it is reasonable or not. > > Option 3) I could also create some "wrapper daemon" in C or other > > language with libvirt binding. But I think that this is an ugly way :-) > > Option 4) I could "talk" directly to the libvirt socket. But I am not > > sure how the communication goes there and whether it is stable or > > changes with every version. And I consider the socket to be internal > > thing of libvirt... > > > > Is there any other option I missed? And which one would you suggest? > > > > Radek > > > > -- > > Libvir-list mailing list > > Libvir-list(a)redhat.com > > https://www.redhat.com/mailman/listinfo/libvir-list

1 0

[libvirt] Need help for thesis on Xen - disabling HAP for benchmarking purposes
by Alexander Sascha 25 Mar '09

25 Mar '09

Hi, I'm a German student, writing my thesis on "Virtualization with Xen. Analysis and Comparision of Different Techniques such as Paravirtualization, Full Virtualization, and Utilization of Hardware Support Provided by the Processor". I'd like to know how I can disable HAP/RVI/Nested Paging for benchmarking purposes. I read somewhere (http://markmail.org/message/bbnivuqx6vjz7jg4) that Xen developers decided to disable the global grub parameter introducing a per domain flag for HAP instead (hap=0/1). I used virt-manager built in CentOS for setting up all domUs; virt-manager uses libvirt so that no "old-style" configuration files are stored in /etc/xen. I actually don't know how if libvirt-XML-files already support something like <hap>0</hap>. How would you proceed to modify an already existing domain using libvirt/virt-manager? I really do apologize for my "newbie ignorance". Thanks in advance for any useful help. - Alex

2 1

[libvirt] [RFC]: Volume allocation progress reporting
by Cole Robinson 25 Mar '09

25 Mar '09

The attached patch implements storage volume allocation progress reporting for file volumes. Currently, the volume creation process looks like: - Grab the pool lock - Fully allocated the volume - 'define' the volume (so it shows up in 'virsh vol-list', etc) - Lookup the volume object to return - Drop the pool lock The new sequence is: - Grab the pool lock - 'define' the volume (even though nothing is on disk yet) - Drop the pool lock - Allocate the volume as needed - Regrab the pool lock - Lookup the volume object to return - Drop the pool lock (again) Since the volume is 'defined', the user can fetch an object reference in a separate thread, and continually poll the 'info' command for up to date numbers. This also has the benefit of dropping the pool lock during the potentially lengthy allocation, as currently 'virsh pool-list' etc. will block while any volume is being allocated. Non file volumes maintain existing behavior. I tried to make the implementation resistant to user error: say if the pool is deactivated or deleted while the volume is being allocated. The creation process may bail out, but I couldn't produce any bad errors (crashing). There are a few other small fixes in this patch: - Refresh volume info when doing volume dumpxml - Update volume capacity when doing a refresh I've also attached an ugly python script that can test this. Presuming you have a pool named 'default', running python sparse.py --vol-create-info Will launch an allocation, print vol.info in a loop. Feedback appreciated. Thanks, Cole import libvirt import threading import time import sys import optparse poolname = "default" volname = "nonsparsetest" testvol = "testvol" uri = "qemu:///system" testvolxml = """ <volume> <name>%s</name> <capacity>1048576000</capacity> <allocation>0</allocation> <target> <format type='raw'/> </target> </volume> """ % testvol volxml = """ <volume> <name>%s</name> <capacity>1048576000</capacity> <allocation>800000000</allocation> <target> <format type='raw'/> </target> </volume> """ % volname failvol = """ <volume> <name>%s</name> <capacity>1048576000</capacity> <allocation>1048576000</allocation> <target> <format type='bochs'/> </target> </volume> """ % volname poolxml = """ <pool type='dir'> <name>default</name> <target> <path>/var/lib/libvirt/images</path> </target> </pool> """ # Helper functions def exception_wrapper(cmd, args): try: cmd(*args) except Exception, e: print str(e) def make_test_vol(): pool = get_pool() pool.createXML(testvolxml, 0) def del_vol(name=volname): pool = get_pool() vol = pool.storageVolLookupByName(name) vol.delete(0) def del_pool(): pool = get_pool() pool.destroy() pool.undefine() def define_pool(): conn = libvirt.open(uri) pool = conn.storagePoolDefineXML(poolxml, 0) pool.create(0) pool.setAutostart(True) def allocate_thread(xml=volxml): try: del_vol() except: pass pool = get_pool() print "creating vol in thread" vol = pool.createXML(xml, 0) print "creating vol complete." def info_thread(vol): for i in range(0, 40): time.sleep(.5) print vol.info() def get_pool(name=poolname): conn = libvirt.open(uri) return conn.storagePoolLookupByName(name) def get_vol(name=volname): pool = get_pool() return pool.storageVolLookupByName(name) def cmd_vol_create(xml=volxml): exception_wrapper(define_pool, ()) pool = get_pool() print pool.listVolumes() t = threading.Thread(target=allocate_thread, name="Allocating", args=(xml,)) t.start() time.sleep(5) print "\nRefreshing pool and dumping list" pool.refresh(0) print pool.listVolumes() def cmd_vol_fail(): cmd_vol_create(failvol) def cmd_vol_poll(): cmd_vol_create() vol = get_vol() t = threading.Thread(target=info_thread, name="Getting info", args=(vol,)) t.start() def main(): import optparse parser = optparse.OptionParser() parser.add_option("", "--vol-create", action="store_true", dest="vol_create", help="Create a volume nonsparse volume that should " "succeed") parser.add_option("", "--vol-create-info", action="store_true", dest="vol_create_info", help="Create a volume nonsparse volume that should " "succeed, and list info in a loop") parser.add_option("", "--vol-create-fail", action="store_true", dest="vol_fail", help="Create a volume that will fail at the allocate " "stage") options, ignore = parser.parse_args() if options.vol_create: cmd_vol_create() elif options.vol_create_info: cmd_vol_poll() elif options.vol_fail: cmd_vol_fail() else: parser.print_help() sys.exit(1) if __name__ == "__main__": main()

2 2

[libvirt] first cut public API for physical host interface configuration
by Laine Stump 25 Mar '09

25 Mar '09

To get started integrating libnetcf support into libvirt, the attached libvirt.h diff has a first attempt at the public API that will hook up to libnetcf on the libvirtd side. I started out with the virNetwork* API, and modified/removed as seemed appropriate. A few points worth mentioning: virNetwork has "defined" and "active" interfaces, but so far virInterface just has interfaces, as the active/inactive status is really controlled by 1) the "onboot" property in the XML, and 2) whether or not virInterfaceStart() has been called yet. The _virInterface struct that is referenced here is more or less identical to _virnetwork. libnetcf works with netcf structs (one per library instance) and netcf_if structs (one per interface, multiple per library instance. As far as I can see right now, the netcf struct will not need to be referenced directly from the client side, but the netcf_if for each interface is needed, and I guess a cookie representing it will be sent to the client side and stored in the _virInterface. Is that what the UUID in _virNetwork is used for? (I know, I know - "read the code!", it's just quicker to (and avoids misreading on my part) to ask) As before, any and all advice/corrections gratefully accepted! diff --git a/include/libvirt/libvirt.h b/include/libvirt/libvirt.h index 779ea72..cac400e 100644 --- a/include/libvirt/libvirt.h +++ b/include/libvirt/libvirt.h @@ -854,7 +854,73 @@ int virNetworkGetAutostart (virNetworkPtr network, int virNetworkSetAutostart (virNetworkPtr network, int autostart); +/* + * Physical host interface configuration API + */ + +/** + * virInterface: + * + * a virInterface is a private structure representing a virtual interface. + */ +typedef struct _virInterface virInterface; + +/** + * virInterfacePtr: + * + * a virInterfacePtr is pointer to a virInterface private structure, this is the + * type used to reference a virtual interface in the API. + */ +typedef virInterface *virInterfacePtr; + +/* + * Get connection from interface. + */ +virConnectPtr virInterfaceGetConnect (virInterfacePtr interface); + +/* + * List defined interfaces + */ +int virConnectNumOfInterfaces (virConnectPtr conn); +int virConnectListInterfaces (virConnectPtr conn, + char **const names, + int maxnames); + +/* + * Lookup interface by name + */ +virInterfacePtr virInterfaceLookupByName (virConnectPtr conn, + const char *name); + const char *uuid); + +/* + * Define interface (or modify existing interface configuration) + */ +virInterfacePtr virInterfaceDefineXML (virConnectPtr conn, + const char *xmlDesc); + +/* + * Delete interface + */ +int virInterfaceUndefine (virInterfacePtr interface); + +/* + * Activate interface (ie call "ifup") + */ +int virInterfaceStart (virInterfacePtr interface); + +/* + * De-activate interface (call "ifdown") + */ +int virInterfaceStop (virInterfacePtr interface); + +/* + * Interface information + */ +const char* virInterfaceGetName (virInterfacePtr interface); +char * virInterfaceGetXMLDesc (virInterfacePtr interface, + int flags); /** * virStoragePool: *

1 0

[libvirt] virt-top
by Zvi Dubitzky 24 Mar '09

24 Mar '09

Are the sources of virt-top available at some repository to look at . I do not see them in the Download section of libvirt Web site thanks Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL

2 1

[libvirt] save/ restore a domain with kvm
by Zvi Dubitzky 24 Mar '09

24 Mar '09

I am working with virsh The version of VIRSH as it shows is : virsh # version Compiled against library: libvir 0.4.4 Using library: libvir 0.4.4 Using API: QEMU 0.4.4 Running hypervisor: QEMU 0.9.1 Under virsh I do the following : 1. suspend a running VM 2. save it to a file with the 'save' command 3. restore the saved domain file with the 'restore' command while the domain is still in the suspend mode or after it was shutdown/destroyed (not seen with 'list --all) In either case I get the error message saying : virsh # restore /home/dubi/xml/vm_dubi2.sav libvir: QEMU error : operation failed: failed to start VM error: Failed to restore domain from /home/dubi/xml/vm_dubi2.xml Any idea what is the failure reason ? The content of the saved domain file isattached: Zvi Dubitzky Virtualization and System Architecture Email:dubi@il.ibm.com IBM Haifa Research Laboratory Phone: +972-4-8296182 Haifa, 31905, ISRAEL

3 5

[libvirt] [PATCH 1/2] VirtualBox support to libvirt
by Pritesh Kothari 24 Mar '09

24 Mar '09

Hi All, I have attached a patch which when applied on the HEAD as of today would allow virtualbox support in libvirt. The patch works very well with the VirtualBox OSE version and the 2.2 Beta release. [PATCH 1/2] contains diff of files already in libvirt. [PATCH 2/2] contains new files needed for VirtualBox support. Regards, -pritesh

3 6

[libvirt] Improve heuristic for default guest architecture
by Soren Hansen 24 Mar '09

24 Mar '09

In libvirt 0.6.1, if you create a domain description of type 'kvm' without an arch set on an x86-64 host, you would get an i686 qemu guest rather than the expected x86-64 kvm guest. This is because virCapabilitiesDefaultGuestArch doesn't take the domain type into consideration, so it just returned the first hvm architecutre that has been registered, which is i686. After applying Dan P's patch, http://www.redhat.com/archives/libvir-list/2009-March/msg00281.html, I now get a i686 kvm guest, since kvm now can do i686 guests from libvirt. This is certainly an improvement, but I think a more reasonable default is to attempt to match the host's architecture. This patch makes virCapabilitiesDefaultGuestArch also check the domain type, and also gives preference to a guest architecture that matches the host's architecture. Index: libvirt-0.6.1/src/capabilities.c =================================================================== --- libvirt-0.6.1.orig/src/capabilities.c 2009-03-19 15:18:09.483317579 +0100 +++ libvirt-0.6.1/src/capabilities.c 2009-03-19 15:42:31.027341187 +0100 @@ -468,14 +468,26 @@ */ extern const char * virCapabilitiesDefaultGuestArch(virCapsPtr caps, - const char *ostype) + const char *ostype, + const char *domain) { - int i; + int i, j; + const char *arch = NULL; for (i = 0 ; i < caps->nguests ; i++) { - if (STREQ(caps->guests[i]->ostype, ostype)) - return caps->guests[i]->arch.name; + if (STREQ(caps->guests[i]->ostype, ostype)) { + for (j = 0 ; j < caps->guests[i]->arch.ndomains ; j++) { + if (STREQ(caps->guests[i]->arch.domains[j]->type, domain)) { + /* Use the first match... */ + if (!arch) + arch = caps->guests[i]->arch.name; + /* ...unless we can match the host's architecture. */ + if (STREQ(caps->guests[i]->arch.name, caps->host.arch)) + return caps->guests[i]->arch.name; + } + } + } } - return NULL; + return arch; } /** Index: libvirt-0.6.1/src/capabilities.h =================================================================== --- libvirt-0.6.1.orig/src/capabilities.h 2009-03-19 15:18:09.507338228 +0100 +++ libvirt-0.6.1/src/capabilities.h 2009-03-19 15:42:31.027341187 +0100 @@ -177,7 +177,8 @@ extern const char * virCapabilitiesDefaultGuestArch(virCapsPtr caps, - const char *ostype); + const char *ostype, + const char *domain); extern const char * virCapabilitiesDefaultGuestMachine(virCapsPtr caps, const char *ostype, Index: libvirt-0.6.1/src/domain_conf.c =================================================================== --- libvirt-0.6.1.orig/src/domain_conf.c 2009-03-19 15:18:09.531341976 +0100 +++ libvirt-0.6.1/src/domain_conf.c 2009-03-19 15:42:31.031345327 +0100 @@ -2146,7 +2146,7 @@ goto error; } } else { - const char *defaultArch = virCapabilitiesDefaultGuestArch(caps, def->os.type); + const char *defaultArch = virCapabilitiesDefaultGuestArch(caps, def->os.type, virDomainVirtTypeToString(def->virtType)); if (defaultArch == NULL) { virDomainReportError(conn, VIR_ERR_INTERNAL_ERROR, _("no supported architecture for os type '%s'"), Index: libvirt-0.6.1/src/xm_internal.c =================================================================== --- libvirt-0.6.1.orig/src/xm_internal.c 2009-03-19 15:18:09.559316828 +0100 +++ libvirt-0.6.1/src/xm_internal.c 2009-03-19 15:42:45.807318313 +0100 @@ -695,7 +695,7 @@ if (!(def->os.type = strdup(hvm ? "hvm" : "xen"))) goto no_memory; - defaultArch = virCapabilitiesDefaultGuestArch(priv->caps, def->os.type); + defaultArch = virCapabilitiesDefaultGuestArch(priv->caps, def->os.type, virDomainVirtTypeToString(def->virtType)); if (defaultArch == NULL) { xenXMError(conn, VIR_ERR_INTERNAL_ERROR, _("no supported architecture for os type '%s'"), -- Soren Hansen | Lead Virtualisation Engineer | Ubuntu Server Team Canonical Ltd. | http://www.ubuntu.com/

3 3

[libvirt] [PATCH] Adding filesystem mount support for openVZ
by Florian Vichot 24 Mar '09

24 Mar '09

Hi everyone, This patch is to allow using the "mount" type in the "filesystem" tag for OpenVZ domains. Example: ... <filesystem type='mount'> <source dir='/path/to/filesystem/directory/' /> <target dir='/path/to/pivot/root/' /> </filesystem> ... This is my first patch to an external project, so don't spare me if I got things wrong :) Also, I'm curious for suggestions as to how I could allow for the target not to be specified in the XML. Because in this case OpenVZ just makes a temporary pivot root in "/var/lib/vz/root/" and that is probably sufficient for most people, who might not want to have to explicitly create a pivot root somewhere, just for mounting the filesystem while it's running. I was thinking either allow for the target tag not to be specified, or add an "auto" attribute to target. Which one sounds better ? Thanks, Florian

3 8