Amazon Server blog

I continue to write articles about Amazon Services and servers here: Amazon Server

Recover broken Amazon EC2 instance

Sometimes shit happens, accidentally.
Especially  when you lost ssh access to your virtual server (e.g. due to errors in  ~/.ssh/authorized_keys, /etc/sudoers, /etc/group, /etc/shadow or another important config file).

When you catch errors like:

No supported authentication methods available

or

sudo
sudo: >>> /etc/sudoers: syntax error near line 1 <<<
sudo: parse error in /etc/sudoers near line 1
sudo: no valid sudoers sources found, quitting
sudo: unable to initialize policy plugin


Don't worry, inside the Amazon cloud you can repair almost all.

Follow these simple steps to solve the problem:

1. Via AWS console:
1. Launch new temporary EC2 micro-instance in the same availability zone.
2. Stop broken instance.
3. Detach root EBS volume from the broken instance.
4. Attach this EBS to the temporary instance (as /dev/sdf).
2. Via SSH console from the temporary instance:
1. Mount this EBS:
   sudo mount /dev/xvdf /mnt
2. Fix all problems on mounted file system.
3. Unmount this EBS:
   sudo umount /mnt
3. Via AWS console:
1. Detach this EBS from the temporary instance.
2. Attach this EBS to the broken instance (as /dev/sda1).
3. Start broken instance.
4. Check that broken instance now is healthy (available via ssh and everything is functioning normally).
5. Then you can stop temporary instance (or terminate it)

I hope this article will make someone little bit happy)

Disk drive for the Amazon virtual server

Amazon virtual server uses ELB (Elastic Block Store) storage as disk drive.

It's like a HDD (Hard Disk Drive) within the physical server.

You can create ELB of any size (1GB..1TB) from the Amazon Web Console.
Then you can attach ELB to EC2 (via Web Console),
format:
sudo mkfs.ext4 /dev/xvdf

mount:
sudo mount /dev/xvdf /mnt/disk1

and use:
sudo mkdir --mode=777 /mnt/disk1/tmp

I guess that internally ELB is just a dynamically expanding file, movable between servers.
And I've expanded ELB to the entire allocated size:

dd if=/dev/zero of=/mnt/disk1/tmp/zero bs=1M

921+0 records in
920+0 records out
965447680 bytes (965 MB) copied, 29.1309 s, 33.1 MB/s


Then ELB is very fast (speed depends on the file size and state of system caches):

(1.0 MB) copied, 0.0018899 s, 555 MB/s
(10 MB) copied, 0.0122498 s, 856 MB/s
(105 MB) copied, 0.442308 s, 237 MB/s
(262 MB) copied, 4.84185 s, 54.1 MB/s