Homework #4 S2005

Goal:

Learn how to configure, patch, and compile User Mode Linux kernel.
Learn how to use Copy-On-Write File System to share the same file system image.
Learn how to configure and run Linux Virtual Server Cluster.

Assignment Date: 4/1/2005
Due Day: 4/7/2005

We are creating a VS-NAT cluster with the following network topology, hostnames, and IP addresse assignment to the network interface:

Note that client, rs1, rs2, and rs3 will be created with User Mode Linux virtual machines. Since we use tuntap for emualting the network interface of the virtual machine, we will need two additional IP addresses for the two tuntap device of each virutal machine. For example, rs1 is configured with eth0=tuntap,,,10.10.10.211 and eth1=tuntap,,,192.168.0.211. The host fc3c will also serve as virtual server or LVS load balancer. Note that you will be asked to answer the following five questions. Therefore read them before you proceed with the exercises.

Answers of the following five questions:

Q1. With "make menuconfig ARCH=um" we select the option/features to be bulit-in the kernel or to be included later as kernel modules.
What symbol indicates the option is to be included as "built-in"? What symbol indicates the option is to be included as "module"? What key is used to select the option as "bulit-in"? What key is used is used for selecting option as "module"?
Q2. For a system with multiple virutal machines, the Copy-on-Write (COW) file systems can reduce the disk space by sharing the same read-only file system. In our case, the rs1, rs2, rs3, and client virtual machines share the same root_fsSW9_sisbak file system image. How many MBytes we have saved by using the COW file system? Hint. The first column of "ls -ls" command show the real size of the file in KB unit.
Q3. What command can be used by the host to mount the file system of the UML virutal machine for preparation or for off-line maintenance?
Q4. In Slackware Linux distribtion,
- a. What file specifies the hostname and will be read by OS during the boot time?
- b. What file specifies the the default IP addresses of the network interfaces and will be read by OS during the boot time?specify
Q5. If we were to add an additional real server, rs4, which has twice the server capacity than those of other real server, what is the vs-nat configuration command(s) we need to add to the vsNATSetup.sh? Make your assumption on the IP address of the real server.

Exercise 1. Configure, Patch, and Compile the User Mode Linux. (about 20-30 minutes)

# Build uml linux kernel
#

# Normally you would download the most recent version of Linux kernel from http://www.kernel.org/
# Then apply the patches
# start fc3c virtual machine on your designated EAS 138 PC.

scp <login>@sanluis.uccs.edu:~cs526/public_html/hw/hw4/linux-2.6.8uml.tar.bz2 .
# here replace <login> with your cs unix login name. Note that after .bz2 there is a space followed by period, which specifies the current directory as the secure copy target directory.

tar xjvf linux-2.6.8uml.tar.bz2
# This creates a linux-2.6.8 kernel source directory and may takes about 8 minutes. You can also use xvf as option for the tar comand, then it will not show the progress with the names of the files being restored but it will speed up the process.

cd linux-2.6.8
# this kernel source directory already being patched with uml patch and ipvs patch.
#
# Apply Julian arp hidden patch

scp <login>@sanluis.uccs.edu:~cs526/public_html/hw/hw4/hidden-2.6.4-1.diff .

patch -p1 < hidden-2.6.4-1.diff
#
# The detailed of the patch result is shown in this file.

scp <login>@sanluis.uccs.edu:~cs526/public_html/hw/hw4/.config .
# This copies a configuration file which I used for kernel compilation.
# We have found that the one comes with linux-2.6.8uml.tar.bz2 does not turn on the IP Virtual Server Configuration for selection.
#
# A fundamental fix is to run "make mrproper" which removes all compiled code and .configuration files, replace with a default configuration
# file. It gives us a "totally" clean start for kernel compilation.

# use make menuconfig to select the features/modules to be included in the kernel

make menuconfig ARCH=um
# the above menuconfig command generates a text-based menu for selecting options and features to be include the kernel or kernel modules.

Here are the list keys for selecting or deselecting the options:

Use Up-arrow and down-arrow to navigate the menu.

Hit key <Enter> selects submenus --->‚

Highlighted letters are hotkeys.

Press key <Y> to include the feature/option,

Press key <N> to exclude the feature/option,

Press key <M> to make include the feature/option as module.

Press <Esc><Esc> to exit, <?> for Help.

Legend: [*] built-in [ ] excluded <M> module < > module capable

We will select options in IP: Virtual Server Configuration and Network Packet filtering submenus.

From the top menu, hit down-arrow to select Networking Support ---> submenu.

Hit enter after the Networking Support menu was selected.

From the Network Support menu, hit down-arrow to select networking option submmeu. Hit enter there.

From the Networking options menu, hit down-arrow about 16 times to select the "IP: Virtual Server Configuration submenu".

From the IP: Virtual Server Configuration submenu, select the following schedulers and leave others unselected as default.

1. round robin

2. weighted round robin

3. least connection

4. weighted least connection

Exit the IP: Virtual Server Configuration menu, by hitting the Esc key.

From the Networking option menu, select the Network packet filtering (replaces ipchains) submenu. It is the two submenu down from the IP: Virtual Server Configuration submenu.

Select IP: Netfilter Configuration submenu. Hit enter.

Select all options here. There are quite a few options there, related to the firewall filtering features on different header field. Our exercise uses NAT and Masquerade.

Conclude the menu selection by hitting Esc keys until it says "Do you wish to save your new kernel configuration?"

Hit enter or Y to confirm the choice

# Compile the UML Linux kernel with the following command. This could take about 11 minutes.

cp linux ../../sis
# copy the linux to ~/sis where the shell scripts of virutal machines, such as rs[n].sh and client.sh will share it.

Exercise 2. Create Copy-On-Write File Systems and Virtual machines. (about 20 minutes)

#copy hw/hw4/lvs configiration files here.
cd ~/sis
scp -r <login>@sanluis.uccs.edu:~cs526/public_html/hw/hw4/lvs/* .
cp lvslab/vs/hosts /etc/hosts
cp lvslab/vs/ipvsadm /sbin

# copy the lvs cluster configuration files in lvslab to the root file system
# to be shared by all virtual machines
mkdir mnt
mount root_fsSW9_sisbak mnt -o loop
cp -r lvslab mnt/root
umount mnt

# If you did not shut down the sis-orig virtual machine properly by "shutdown -h now", the root_fsSW9_sisbak root file system may be in a
# inconsisten state that needs to be repaired. Instead repairing on each real server virtual machine, it is a good idea to run ./sis.sh to make # sure the shared file system is OK. You can also check if the lvslab directory is copied properly in the above command.
./sis.sh

# You will see the file sytem being patched during the startup and the login prompt with "sis-orig:" You will also see a x terminal login shell window appear similar to your hw2.
# Just enter root loginname, password, and then enter "shudown -h now". Now your root shared file system is fixed.
# Make sure you shutdown this virtual machine before you make the COW file systems in the next step.
# Otherwise the COW file system will be garbled. From this point forward, we cannot change the content of the root_fsSW9_sisbak.

# Create copy-on-write (COW) file systems for each virtual machines using the following script.
# Make sure your understand the syntx of the commands contained in all the scripts of these exercises.
./mkcow.sh
# check the size of the copy on write file systems
ls -ls
# record the actual size of the cow file system (in KB).

# We could have mount the cow file system and modify the system configuration
# file of each virtual machine before actually start it.
# It will avoid having to reboot for hostname to take effect.
# However when I use "mount root_fs_cow_rs1 mnt -o loop" it asks the file system type.
# I added "-t ext2" and it said "mount: wrong fs type, bad option, bad superblock on /dev/loop0
# Look like the newly created cow file system missing some file system info and can not be recognized by
# the mount command.

# start rs1 virtual machine
# open anotehr terminal window
cd sis
./rs1.sh
# you will see the following prompt:
# welcome to 2.6.8.1-lum (tty0)
# sis-orig login:
# another virtual Console will open with (rs1) as label
# login as root, same password
#
cd lvslab/rs1
./rsSetup.sh
reboot
#
# During reboot, if you see the following error msgs:
#
# VFS: Cannot open root device "98:0" or unknown-block(98,0)
# Please append a correct "root=" boot option
# Kernel panic: VFS: Unable to mount root fs on unknown-block(98,0)
#
# This is due to the root file system has not been released by the
# host OS
# after reboot the HOSTNAME will take effect and you will see
# rs1 login:
# login again
#
# check the network interface
ifconfig
# you should see eth0 set with 10.10.10.111
# and eth1 set with 192.168.0.111
#
# check the routing table
netstat -rn
# you should see the default gw set to 10.10.10.1
#
# go to the lvs configuration directory
cd lvslab/rs1
#
# on host machine, start the browser and
# type "http://rs1/
# you should see rs1.csnet default web page with pink background
#

# Repeat the above steps for setting up rs2 virtual machine
# substitute "./rs1.sh" with "./rs2.sh"
# substitute "cd lvslab/rs1" with "cd lvslab/rs2"
# substitute "http://rs1/" with "http://rs2"
# substitute "cd lvslab/rs1" with "cd lvslab/rs2"
#

# Repeat the above steps for setting up rs3 virtual machine
# substitute "./rs1.sh" with "./rs3.sh"
# substitute "cd lvslab/rs1" with "cd lvslab/rs3"
# substitute "http://rs1/" with "http://rs3"
# substitute "cd lvslab/rs1" with "cd lvslab/rs3"
#

# Repeat the above steps for setting up client virtual machine
# substitute "./rs1.sh" with "./client.sh"
# substitute "cd lvslab/rs1" with "cd lvslab/client"
# Here we do not have to check the web sever on client
#

Exercise 3. Configure LVS-NAT Cluster (5 minutes)

# on each of the rs servers, run the rsNATSetup.sh at its lvslab/rs[n] configuration directory. For example, on rs1,
cd lvslab/rs1
./rsNATSetup.sh
# on vs (host),
cd ~/sis/lvs/lvslab/vs
./vsSetup.sh
./vsNATSetup.sh
# Disable the firewall of the host to allow client http request to pass through. Type the following command to set firewall configuration.
lokkit
# You should the following text-based dialog window.

# Hit tab key to switch among options. Hit space bar to confirm the selection.
# Hit tab key to switch to "Disabled". Hit space bar to select "(firewall) Disabled" as security level.
# Hit tab key to switch to "OK". Hit enter to confirm the setting.
test the cluster by
# type "lynx http://vip1/" on the client console.
# observe the web page that identifies the real server that is scheduled by the load balancer to serve the web request.
# hit q to exit from the lynx.
# repeat the "lynx http://vip1/" (by hitting the up arrow to bring back the command and then hit enter)
# and observe the web pages provided by the different real servers
# see if it follows the wrr scheduling discipline
#
Try http://vip2/" and observe the different pattern.
Experiment with different weights on real server by modifying vsNATSetup.sh
Capture the image of the screen with vs, rs1, rs2, rs3, and client similar to the one below.
Copy the screendump image as vsnatcluster.png and save it in you web site on CS Unix machine.
Email me the following as your hw4 submission.
- the label of machine you used, such as eas-138-20.
- your observation on the access pattern for the cluster vip1 and vip2,
- the url of the screendump (make sure your verify the image shows up on a browser.
- Answers of the following five questions:
- Q1. With "make menuconfig ARCH=um" we select the option/features to be bulit-in the kernel or to be included later as kernel modules.
  What symbol indicates the option is to be included as "built-in"? What symbol indicates the option is to be included as "module"? What key is used to select the option as "bulit-in"? What key is used is used for selecting option as "module"?
- Q2. For a system with multiple virutal machines, the Copy-on-Write (COW) file systems can reduce the disk space by sharing the same read-only file system. In our case, the rs1, rs2, rs3, and client virtual machines share the same root_fsSW9_sisbak file system image. How many MBytes we have saved by using the COW file system? Hint. The first column of "ls -ls" command show the real size of the file in KB unit.
- Q3. What command can be used by the host to mount the file system of the UML virutal machine for preparation or for off-line maintenance?
- Q4. In Slackware Linux distribtion,
  - a. What file specifies the hostname and will be read by OS during the boot time?
  - b. What file specifies the the default IP addresses of the network interfaces and will be read by OS during the boot time?specify
- Q5. If we were to add an additional real server, rs4, which has twice the server capacity than those of other real server, what is the vs-nat configuration command(s) we need to add to the vsNATSetup.sh? Make your assumption on the IP address of the real server.

Homework #4. Compile UML Linux and Configure a Linux Virtual Server Cluster