Проект

Общее

Профиль

OPERATIONS » История » Версия 2

« Предыдущее - Версия 2/23 (Разница(diff)) - Следующее » - Текущая версия
Владимир Ипатов, 24.10.2012 18:16


OPERATIONS

Cluster management
Instance management
Puppet
DNS&

Virtualization and HA

SCI cluster virtualization and HA capabilities are really bringed
by a Ganety installation over the Debia/GNU Linux, so you should
read Ganeti manuals on how to operate with virtualisation and HA
features.

The main documentation page: http://docs.ganeti.org/ganeti/master/html/
A brief operations overview: http://docs.ganeti.org/ganeti/master/html/walkthrough.html

Ganeti operations should be issued on the 'master' node which has additional the
cluster's IP as an interface alias.
The first new nodes becames the 'master candidates', so they can be promoted to master
if the master node will fail.

Major commands:

Check if all is ok

gnt-cluster verify

Show detailed information

gnt-(cluster|node|instance) info

List objects

Nodes

gnt-node list

Configured instances

gnt-instance list

Create new instance

New instance on the NODE1 with failover on the NODE2, with one disk size 10Gb and 256M RAM.
INSTANCE_NAME must be resolvable via DNS.

gnt-instance add -t drbd -o debootstrap+default -s 10g -B memory=256m -n NODE1:NODE2 INSTANCE_NAME

Real example: create new instance
  • on the node gnt2
  • with failover on gnt1
  • with three disks:
    • 10g on volume group xenvg
    • 600m on volume group xenvg (probably for swap)
  • 5g on volume group xenvg
  • DNS name vg-test1
  • IP address will be configured automatically by resolving vg-test1' IP and matching it against /etc/ganeti/networks
gnt-instance add -t drbd -o debootstrap+default --disk 0:size=10g,vg=xenvg --disk 1:size=600m,vg=xenvg --disk 2:size=5g,vg=newvg -B memory=256m -n gnt2:gnt1 vg-test1

NOTE: the vg=.... option to specify different volume groups will be available since Ganeti 2.4.0 release.

Instance manipulations

Various simple instance manipulations

gnt-instance (startup|shutdown|reboot|failover|remove|console) INSTANCE

OS command execution

Run OS command on all nodes

gnt-cluster command .....

SSH

All new instances, including 'sci' are going with empty root password and no ssh server installed.
You should first log in to the instance using gnt-instance console INSTANCE command, then set
root's password using passwd command and install ssh server using

apt-get install openssh-server

PUPPET

Puppet http://www.puppetlabs.com/) is the open source platform for enterprise systems management.
Puppet is used to deploy the configuration to the cluster instances.
Documentation can be found at http://docs.puppetlabs.com/

Puppet master is installed on the instance 'sci'.

DNS

DNS is configured on the instance 'sci' via puppet.

The zone file is stored as a template at /etc/puppet/modules/bind9/templates/sci/zone.erb
On each update you should set new (growing) zone serial number, initially

0000000001 ; Serial

to the new growing value. YYYYMMDDNN is recommended (NN - is the change number in one day).
Feel free to modify it to adjust your system.

New names should be placed below the string

; here you can put any other records

To apply your changes quickly you should issue

/etc/init.d/puppet restart

Apt

Apt is configured on the instance 'sci' via puppet.

sources.list

Global distributed sources.list template is resided in /etc/puppet/modules/approx/templates/sources.list.erb

Approx

Approx is configured in /etc/puppet/modules/approx/templates/approx.conf.erb
To apply your changes quickly you should issue

/etc/init.d/puppet restart