Xen Install. Mini how-to.
Contents
1. what and currently enable most of them I have to spend some time on this issue 6 2 tc My experience in traffic shaping with tc is based on the following how to http linux net osdl org index php Netem I still don t know whether tc evolved enough to support ingress shaping if not we have to adopt Kirk s solution in dom0 I assume that the good way to shape traffic between VMSs is to apply tc rules to the virtual interfaces exported to VMs After VM boot they always take form of vif lt vm id gt lt interface number gt 7 Related resources Xen at Clarkson http xen cosi clarkson edu Xen User s Manual http www cl cam ac uk Research SRG netos xen readmes user user html Mercurial hg Cheatsheet for Xen http www xensource com xen documentation hg cheatsheet txt Install Guest Image http wiki xensource com xenwiki InstallGuestImage 8 References 1 xen devel mailing list 20 Jul 2005 Xen devel Xen Dell 2850 PERC 4e Di lock up http lists xensource com archives html xen devel 2005 07 msg00659 html 2 Xen Bugzilla database fixed 2005 08 29 http bugzilla xensource com bugzilla show_bug cgi id 76 102. Xen Install Mini how to December 8 2005 1 Introduction 2 Getting the latest Xen sources Xen uses Mercurial hg repository Here is an original how to from Xen team describing how to use it is here http www xensource com xen documentation hg cheatsheet txt However actually you need only three steps to download latest version of Xen Initialize repository hg init Configure the repository root parent by putting the following into hg hgrc For Xen 3 0 un stable paths default http xenbits xensource com xen unstable hg or for Xen 2 0 paths default http xenbits xensource com xen 2 0 hg Pull the most recent changes hg pull u 2 1 Compiling Xen Root directory of the Xen tree contains README file describing how to compile Xen However most of the time you just have to run make world make install This will put Xen related configuration files into etc xen and install Xen dom0 and domU kernels into boot directories Further information about build process is available in the README file in root directory of Xen source tree Note that the README file recommends to build Xen with the following command make KERNELS linux 2 6 xen world README file says that it will build Xen kernel capable to run in both dom0 and domU instead of two separate kernels Unfortunately I believe it contains an error and in contrast to README omitting KERNELS linux 2 6 xen argument to make install you should in
3. asic bridge creation and configuration Thus the following command creates the xen br routeA bridge etc xen scripts network bridge start bridge xen br1 netdev eth5 antispoof no Later you can export this bridge to Xen VM by mentioning it in configuration file see section 4 3 5 Setting up the virtual LAN 5 1 Recompiling your kernel to use VLAN In order to use VLANs you have to recompile your dom0 kernel with 802 1q enabled Like for any other Linux kernel option there are two ways to do that either through the GUI configuration environment or by editing config file In first case you have to run make linux 2 6 xen0 config CONFIGMODE menuconfig And following the path Device drivers gt Networking support gt Networking options gt 802 1Q gt VLAN Support enable 802 1Q support Alternatively you can directly edit Linux kernel configuration file config in linux 2 6 12 xen0 directory and set CONFIG_VLAN_8021Q m what means compile 802 1Q as a module In both cases you have to recompile and reinstall your kernel by invoking make linux 2 6 xen build make linux 2 6 xen install 5 2 Configuring the virtual LAN The following four commands create a bridge attached to the VLAN with id 573 vconfig add eth2 573 ifconfig eth2 573 up etc xen scripts network bridge start bridge xen br1 netdev eth2 573 antispoof no bretl addif xen br1 eth2 573 First command creates VLAN device on the eth2 interface VLAN device name wi
4. ll be eth2 573 Next command brings it up after which brctl bridge is created on this device the help of Xen network bridge script Note what is strange Xen network bridge script doesn t attach device eth2 573 to the new bridge therefore fourth command attaches the bridge explicitly 5 3 Checksum offloading bug In order to improve network I O performance Xen drivers allow VMs to use checksum offloading assuming that when packet is be transmitted between VMs on the same physical host actual checksum need not to be calculated In case when packet lives physical node Xen driver passes packet to actual hardware which calculates checksum Unfortunately this fails for packets transmitted over VLAN In this case checksum will not be calculated and receiver on the other side of the VLAN will drop the packet In order to prevent that I explicitly disable checksum offloading inside VM by invoking 8 ethtool K eth0 tx off However the better solution will be to either recompile Linux kernel or configure the system to not offload the checksum 6 Traffic shaping 6 1 Recompiling your kernel In order to use traffic shaping you also have to change configuration and recompile dom0 Linux kernel see section 5 1 for details Do do that modify Device drivers gt Networking support gt Networking options gt QoS and or fair queueing and enable bunch of related options unfortunately I don t have exact knowledge of which option means
5. ll create LVM device on the file To do that create large file for LVM device dd if dev zero of lvm img bs 1M count 1 seek 3072 Mount it via a loopback device losetup dev loop1 lvm img Initialize partition to support LVM volumes pvcreate dev loop1 Create an LVM volume group called vg vgereate vg dev loopl Create a logical volume called myvmdisk1 of size 2500MB lvereate L2500M n myvmdisk1 vg Now you should now see dev vg myvmdisk1 device Make a filesystem mount it and populate it e g mkfs t ext3 dev vg myvmdisk1 Copy root partition on LVM volume and create necessary directories on it mount dev vg myvmdisk1 mnt guest mount dev sda2 mnt host cp ax mnt host root dev var etc usr bin sbin lib mnt guest mkdir mnt guest proc sys home tmp Unmount everything umount mnt guest umount mnt host Don t forget to edit fstab file and do other changes to guest VM image Create a couple of 100MB copy on write clones lvcreate s L100M n myclonedisk1 dev vg myvmdisk1 lvcreate s L100M n myclonedisk2 dev vg myvmdisk1 5 Now you are able to export LVM COW disk to VM as disk phy dev vg myclonedisk1 sdal w Note that in this example original LVM image was pretty small only 3GB thus we created very small only 100MB COW shadow images We have to think about how large should be the shadows in Emulab Note that you can enlarge the shadow by invoking lvextend 100M dev vg myclonedisk1 However a
6. ll shadows and parent images should fit into original LVM 3GB image 4 3 Configuration files In order to start Xen VM you have to create configuration file for it The simple configuration file can look like kernel boot vmlinuz 2 6 12 xenU memory 128 name VM1 cpu 1i disk phy dev loop1 sda1 w vif bridge xen br routeA bridge xen br routeAB root dev sdai ro ramdisk tmp sda4 images initrd fc3 img extra ro selinux 0 3 Here VM is configured to have 128MB of memory and emulate single CPU machine The physical device dev loop1 will be exported to VM as sdal device Two virtual interfaces bridge xen br routeA bridge xen br routeAB will be also exported to VM usually as well as in the example above interfaces exported to VM are Linux bridges Note that VM name should be unique initrd is passed to Fedora Core which requires it to boot Other quest OSes don t need it see section 2 2 for how to create initrd ram image Note that you can use vif variable to specify MAC address for the VM s NIC vif mac 00 16 3E F6 BB B3 A couple of example configuration files installed with Xen VMM can be found in etc xen Note Xen documentation says you can export a filesystem to domU directly via a file containing this filesystem Thus if you have a disk image full disk with partition table partitions boot record etc you can pass it as disk file tmp sda4 images r
7. outer_A img sda w Or if you have an image containing only one partition 6 disk file tmp sda4 images sdal img sdal w Unfortunately I wasn t able to figure out the way how to do that correctly both disk and partition images either hang domU on boot as in case of disk image domU is unable to find root on the disk or report that virtual block device vbd cannot be connected as in case of single slice image For more information on that issue look Section 6 2 Using File backed VBDs in Xen User s manual http www cl cam ac uk Research SRG netos xen readmes user user html 4 4 Starting VM In order to start VM you have to invoke the following command xm create lt configuration file gt In order to access console of the started VM invoke xm console lt VM id gt Where lt VM id gt can be learned by running xm list Alternatively you can pass c option to xm upon VM creation and connect to VM s console immediately after start 4 4 1 lib tls bug Upon the boot both dom0 and domU complain that tls emulation is slow and recommend to disable tls by renaming system library i e mv lib tls lib tls disabled In my case this step hasn t made any effect however I remember that it worked for Xen 2 0 That s strange and I have to spend some time figuring what s going on 4 5 Setting up the bridge Although you can create and configure bridge by yourself look man brctl for information Xen provides scripts for b
8. re however you can generate the new initrd ram image by invoking the following commands depmod 2 6 12 6 xen mkinitrd v f with aacraid with sd_mod with scsi_mod initrd 2 6 12 6 xen img 2 6 12 6 xen You can find some additional information in the README file from the root of Xen source directory 3 Starting VM Environment After rebooting into Xen you have to start xend Xen control daemon by invoking xend start You can configure xend to start certain domains automatically look Section 3 3 Starting and Stopping Domains Automatically in the Xen User s Manual http www cl cam ac uk Research SRG netos xen readmes user user html for how to do that After starting xend you can configure and start VMs Xen domUs 3 1 Default gateway bug Currently xend script has the following error on startup it destroys default route to 155 98 36 1 control router emulab net what efficiently makes Emulab node unreachable as well as brakes your connection So far I invoke it as follows xend start route add default gw 155 98 36 1 I hoped that Xen team will fix this bug and haven t payed attention to it but now it s time to understand what s wrong 4 Starting VMs In order to run the simplest VM you have to create a disk image to boot from Note that you can boot VM from your physical partition in read only or even in write mode however it doesn t make much sense for us Therefore I will consider only situations when yo
9. u plan to start Xen VMs from separate disk images 4 1 Creating Xen bootable image The following commands were taken from the following how to unfortunately it was modified not in a good way recently http wiki xensource com xenwiki InstallGuest Image 4 1 1 Creating slice image Create image file here I assume 3GB file dd if dev zero of hd img bs 1M count 1 seek 3072 Associate a loopback device with the file losetup dev loop1 hd img Install filesystem mkfs ext3 dev loop1 Mount host operating system and new disk mkdir p mnt host mkdir p mnt guest mount dev sda2 mnt host mount dev loopl mnt guest Copy files and create necessary directories cp ax mnt host root dev var etc usr bin sbin lib mnt guest mkdir mnt guest proc sys home tmp Unmount host umount mnt host Edit etc fstab to look something like dev sdal ext3 defaults 11 dev devpts dev pts devpts gid 5 mode 620 00 dev shm dev shm tmpfs defaults 00 dev proc proc proc defaults 00 dev sys sys sysfs defaults 00 dev sda3 swap swap defaults 00 dev fd0 media floppy auto pamconsole exec noauto managed 0 0 dev hdc media cdrom auto pamconsole exec noauto managed 0 0 Unmount image file umount mnt guest losetup d dev loop1 Now you are able to pass this file as a partition image to Xen VM mounting it via a loopback device losetup dev loop1 hd img 4 1 2 Creating disk image 4 2 Using LVM As in the previous sections we wi
10. voke both commands with this argument otherwise it will not install your kernel make KERNELS linux 2 6 xen world make KERNELS linux 2 6 xen install Currently kernel built in this way hangs on boot So far Im unable to understand what s wrong Therefore we have to start from the two kernel configuration Another argument for doing that is optimization For us it s probably better to have separate kernels optimized for domU and dom0 Note that you built one kernel you have to adjust dom0 kernel name in grub conf accordingly Note do not try to build such large project in NFS build will most probably fail 2 2 Configuring GRUB Xen uses GRUB loader for booting Thus you have at first install GRUB I can write about that later and then edit GRUB configuration file boot grub grub conf to look like default 0 timeout 5 We can control GRUB via COM1 Disable by default as it hangs on systems without a serial port serial unit 0 speed 115200 word 8 parity no stop 1 terminal timeout 5 serial console title Xen 3 0 XenoLinux 2 6 kernel boot xen 3 gz dom0_mem 131072 com1 115200 8n1 noirqbalance module boot vmlinuz 2 6 xen0 root dev sda2 ro console ttyS0 module boot initrd 2 6 12 1 1390_FC4 emulab 1 img Parameter noirqbalance above is a hack to prevent Dell xxxx box from rebooting see 1 2 for details Note also that I used initrd 2 6 12 1 1390_FC4 emulab 1 img from Emulab s installation of Fedora Co
Download Pdf Manuals
Related Search