system rescue basic

1. add init=/bin/sh to grub or lilo to quickly bootup system , but ctrl-c or ctrl-z might not work
2. mount -t proc none /proc to get command like ps to work, or anything else use /proc file system
3. mount -o remount,rw /
4 .fuser -am /mnt/muic
fuser -km /mnt/music
ps axwwl | grep processidnumber
lsof /mnt/music

5. dd if=/dev/hda of=filename bs=512 count=1
restore partition table:
dd if=filename of=/dev/hda bs=1 count=64 skip=446 seek=446

to cloning bad disk
dd if=/dev/hda of=/dev/hdb conv=noerror,sync
ddrescue --max-retries=-1 /dev/hda /dev/hdb

6. recover lost partition
gpart /dev/hdb (guess partition table)
gpart -b filename -W /dev/hda /dev/hda
note: scan /dev/hda ,backup the existing MBR, then write to /dev/hda

7. in case the bad block on hard disk is in journal file system, so remove journal might help
tune2fs -f -O ^has_journal /dev/hda1
tune2fs: attempt to read block from file system resulted in short read while reading journal inode

# debugfs /dev/hda1
debugfs: features
defugfs : open /dev/hda1
debugfs: quit

Cloning a CD under Linux command line

Jephe Wu(linuxtechres.blogspot.com)

It's easy to clone a standard CD under Linux command line, the following are the steps:

1. find out the volume size (assuming your cdrom is /dev/hdc)
# isoinfo -d -i /dev/hdc
record down volume size (e.g. 53488)

2. generate md5sum from CD
# dd if=/dev/hdc bs=2048 count=53488 | md5sum

3. generating iso file from the existing CD
# dd if=/dev/hdc of=/linuxtechres.iso bs=2048 count=53488 conv=notrunc

4. generate md5sum from iso file
# md5sum /linuxtechres.iso
make sure the output is the same as the step 2

5. burn the iso file

# cdrecord dev=/dev/hdc speed=10 padsize=63s -pad -dao -v -eject /linuxtechres.iso

note:
a. always use padsize=63s -pad -dao to generate a CDR which can be used anywhere.
b. do not use the maximum speed your cd writer can support, use a bit lower speed to write.

6. verify the CD
a. put in CDR, run command 'isoinfo -d -i /dev/hdc' to find out volume size value, let's say it's 53488
b. dd if=/dev/hdc bs=2048 count=53488 | md5sum
make sure the output is the same as step 2 and 4.

That's it.

Cloning a Linux server - general method

Jephe Wu http://linuxtechres.blogspot.com


Last time, we talked about using dd to do Linux server cloning for the same kind of hardwares - Cloning a Linux server - scenery 1. This time, we will introduce another general method for doing Linux server cloning.

I assume the source is using the normal partition which means it's not using software raid, LVM etc. For cloning Linux server with LVM partition, please check my another article at http://linuxtechres.blogspot.com/2007/08/clone-linux-server-with-lvm2-partition.html . My test environment is as follows:

the source:
a. HP DL380 (Intel Xeon CPU)
b. Redhat Enterprise Linux 4

the destination:
a. HP DL385G2 AMD64

The following are the cloning steps:

1. boot up the destination server from RIP(Recovery Is Possible) CD
You can choose the second options in GRUB bootup menu to load everything into memory and skip keyboard map option, then you can eject CD out before doing the following if you'd like to.

2. configure the IP address for the destination server( we use 192.168.0.101 here)
ifconfig eth0 192.168.0.101 up

2. copy over the 3 files from the source(assuming server name is linuxtechres, the IP is 192.168.0.100)
scp 192.168.0.100:/etc/passwd /etc/
scp 192.168.0.100:/etc/shadow /etc/
scp 192.168.0.100:/etc/group /etc/

3. clone the partition table over using sfdisk
Assuming /dev/cciss/c0d0p1 is / partition, /dev/cciss/c0d0p2 is the /boot partition.

ssh 192.168.0.100 'sfdisk -d /dev/cciss/c0d0' | sfdisk /dev/cciss/c0d0
mkfs -t ext3 /dev/cciss/c0d0p1
e2label /dev/cciss/c0d0p1 /
mkfs -t ext3 /dev/cciss/c0d0p2
e2label /dev/cciss/c0d0p2 /boot
Note: check your source server /etc/fstab so that you know why we use e2label to label the partition / and /boot. That's the default label name by Redhat Enterprise Linux for / and /boot partitions.
cd /mnt
mount /dev/cciss/c0d0p1 hd
cd hd
mkdir boot proc sys mnt ...
mount /dev/cciss/c0d0p2 boot
note: if you have some more partitions like /usr and /home etc, make sure you make file system for all partitions and give the correct label as what the source has, similar as the above steps.

4. clone everything over now
cd /mnt/hd
ssh 192.168.0.100 'cd /; tar --exclude ./proc --exclude ./sys --exclude ./mnt -cpf - .' | tar xvpf -
note: you might want to exclude some data folder or partitions also so that you can finish the cloning process earlier, after that, you can copy the data folder/partition separately. see the step 8
cd etc
rm -f blkid.tab mtab (because they will be generated by OS each time it boots up)
vi sysconfig/network-scripts/ifcfg-eth* ( to comment out hardware address lines and change ip address line to the destination IP)

5. install GRUB (if both the source and the destination are x86 32bit)
cd /mnt/hd
chroot .; grub_install hd0
In my case, I skiped this step and use step 7 to install GRUB since the destination is AMD64 CPU

6. umount those partition that you mounted before
exit (exit from chroot environment)
cd /mnt/hd
umount boot mnt
cd /mnt
umount hd
sync;sync;sync
reboot

7. install GRUB on AMD64 servers
If you are cloning from x86_32 server to x86_64 server like AMD64, you have to
boot from the first x86_64 RedHat Enterprise Linux 4 CD to install GRUB
chroot /mnt/sysimage;grub_install hd0

8. rsync over the data partitions
after the OS boots up, you can use rsync to copy the data folder or partition over.
rsync -e ssh -P -avz source_ip_addr:/data/ /data/

That's it. Now you can bring up one more server which is exactly the same as the source.

Cloning a Linux server - use dd

Jephe Wu http://linuxtechres.blogspot.com

One of the benefits of using Linux is that you don't have to install it every time, you can ALWAYS clone it despite the destination server hardware is the same as the source ones or not. There're two situations as follows:

1) exactly same hardwares.
2) different hardwares

For each situation, there're different ways to clone the servers. Today, I'll introduce how to clone server by using command dd.

In order to use dd. Generally, the destination server will be having the same kind of server model as the source, or at least the same kind of network card and hard disk.

dd will clone everything from source to destination, sector by sector. dd doesn't care about what kind of OS, it can apply to Windows too. The advantage is that you only need to use one command, the disadvantage is that it might take a bit longer since it will copy every sector of your hard disk, even there's no data on those sectors. If the time is not important to you, and the hardware is the almost same, it might be the best choice for you to clone servers.

The following is the steps I use.

a. keep the minimum application running on the source if possible.
Shutdown those services as long as you can. The idea here is to prevent anything from writing things to the hard disk during cloning process. However, according to my experience, it works well even if you don't shutdown anything, just keep the source server running as normal while cloning.

b. use RIP(Recovery is possible) CD to boot up the destination server.
You can download RIP CD from here. The smaller non-X version is enough, around 37M. Burn it to CD, boot up the destination server.

Normally, I choose the second GRUB bootup option for RIP so that I can eject out the CD after boot up, everything will be in memory only.

c. after booting up RIP, login as root(no password is needed), running the following commands to set up IP address:
ifconfig eth0 ip_address_here up (IP address here should be the same network segment as the source)

d. run the following commands to clone:
ssh source_server_ip 'dd if=/dev/sda' | dd of=/dev/sda

note: if the hard disk is HP hardware raid, it might be /dev/cciss/c0d0. Just replace it with yours.

Now you have to wait until it finish.

Note: the dd won't give you any process report as long as the things go well. However, you can use the command

pkill -USR1 ^dd$

to get the progress report, for more info, you can refer to this excellent article at
http://www.redhatmagazine.com/2007/08/16/tips-from-an-rhce-how-can-i-make-dd-give-me-a-progress-report/
(How to make dd to give me progress report)

If you want to check remotely if it finish or not. Then you can run the following commands after setting up ip on destination server and before doing actual cloning:

passwd root (to change root password)
/usr/sbin/sshd (to start sshd daemon for remote login)

You can also boot up the source using RIP CD, then use the above commands to change root password for the source and start sshd daemon for you login from the destination server.

That's it. I'll introduce other methods for cloning servers for same kind of hardwares later.