wiki:InstallationGuidelines/Amazon

Version 15 (modified by Praneeth Bodduluri, 10 years ago) ( diff )

--

Amazon EC2

Amazon's Cloud provides a flexible platform to deploy Eden scalably.

The costs aren't fixed & can be difficult to predict, despite their calculator, but are competitive, especially in Singapore, which is a good base for the Asia Pacific region.

Regions & Zones

Amazon supports multiple Regions in order to provide a service closest to your users.

  • Namespaces of Instances, Volumnes & Snapshots are unique only within a Region.
  • Within each Region, there are a couple of Availability Zones to allow spreading the risk across different facilities.
  • Volumes are located within a specific Availability Zone
  • Bandwidth transfers are free within an Availability Zone

Instance Size

  • The free starter 'micro' instance is flexible as it can run both 32-bit & 64-bit Operating Systems.
  • The normal production 'small' instance can only run 32-bit.
  • Larger production instances can only run 64-bit, so can't have the exact same image used.
  • The community Debian Squeeze AMI seems a fine base & attached scripts turn this into an Eden instance

Instance Persistence

  • EBS-backed instances have persistent storage even whilst powered down, which is very useful.
    • For DB I/O performance increase can stripe multiple EBS
      • monitoring data available to see if this is the issue
  • Each time you start an instance up, it will be assigned a new IP ('Public DNS') although this can be overcome using an Elastic IP

Authentication

SSH

This can provide an early stumbling block.

  • Each instance created needs to start with a unique SSH keypair
  • When setting up an instance, be sure to safely download the private key.
  • In order to get the public key (needed by SecureCRT for instance) then you need to login using CLI & retireve it:
    ssh -l root -i private.pem <hostname>
    cat ~/.ssh/authorized_keys
    

SecureCRT needs the private key storing as <filename> & the public as <filename.pub> (all on one line)

SSH will also need to be enabled in your Firewall policy (restrict the IP source, if possible)

Building AMIs for easier deployment

This will ease deployment for others.

  • Choose a debian AMI in your desired ZONE with 1 GB of EBS storage [This enables AMI registration in AWS free tier.]
  • Launch the instance.
  • Copy the installation and configuration scripts -> install-eden-cherokee-postgis.sh and configure-eden-cherokee-postgis.sh into the launched instance.
    wget http://eden.sahanafoundation.org/raw-attachment/wiki/InstallationGuidelines/Linux/Server/CherokeePostgreSQL/install-eden-cherokee-postgis.sh
    chmod a+x install-eden-cherokee-postgis.sh
    wget http://eden.sahanafoundation.org/raw-attachment/wiki/InstallationGuidelines/Linux/Server/CherokeePostgreSQL/configure-eden-cherokee-postgis.sh
    chmod a+x configure-eden-cherokee-postgis.sh
    
  • Run the install-eden-cherokee-postgis.sh script and delete it. [Note: This step takes about 10min - grab a coffee]
    ./install-eden-cherokee-postgis.sh
    rm install-eden-cherokee-postgis.sh
    

CLI

To use any of the AWS CLI tools on your own machine to remotely manage instances, then you need to generate a unique X.509 Certificate per account. This can be done from the 'Security Credentials' page within your account.

CLI Management

There are extensive CLI tools available to manipulate your instances.

Growing Storage

1Gb EBS is too small for Eden - need to grow to 3Gb

# --region ap-southeast-1
set EC2_URL=https://ec2.ap-southeast-1.amazonaws.com
set ZONE=ap-southeast-1b
set DEV=i-9546edc0
# Stop Host
ec2stop %DEV%
# Create a snapshot
ec2-create-snapshot vol-e189e88c
# Create new volume from snapshot
ec2-create-volume -z %ZONE% --size 3 --snapshot snap-63f89d08
# Delete Snapshot (if no data in yet)
ec2-delete-snapshot snap-63f89d08
# Attach new volume as secondary
ec2-attach-volume -i %DEV% vol-a9c2a3c4 -d /dev/sdb1
# Start Host
ec2start %DEV%
ec2-describe-instances
# Login (Remember different IP!)
mkdir /mnt/data
echo '/dev/xvdb1 /mnt/data ext3 defaults,noatime 0 0' >> /etc/fstab
mount /mnt/data
resize2fs /dev/xvdb1
umount /mnt/data
# Stop Host
ec2stop i-c75af292
# Unattach volumes
ec2-detach-volume -i %DEV% vol-e189e88c
ec2-detach-volume -i %DEV% vol-a9c2a3c4
# Delete old volume
ec2-delete-volume vol-e189e88c
# Attach volume as boot
ec2-attach-volume -i %DEV% vol-a9c2a3c4 -d /dev/sda1
# Attach old volume for /var/log
ec2-attach-volume -i %DEV% vol-e189e88c -d /dev/sdb1
# Start Host
ec2start %DEV%
ec2-describe-instances
# Login (Remember different IP!)
df -h
# Use the old partition for /var/log (to avoid DoS)
mv /var/log /var/log_old
vim /etc/fstab
/dev/xvdb1 /var/log  ext3    noatime 0 0

mount /var/log
mv /var/log_old/* /var/log
rm -rf /var/log_old

Keep Templates as EBS Volumes as this is cheaper than Snapshots

Attachments (14)

Note: See TracWiki for help on using the wiki.