Decommissioning Servers

From Wikitech
(Difference between revisions)
Jump to: navigation, search
Line 4: Line 4:
 
* If database server, ensure it is not online in any cluster and that mysql services are not running (or slaved to another server, or providing the master server for others to replicate.)
 
* If database server, ensure it is not online in any cluster and that mysql services are not running (or slaved to another server, or providing the master server for others to replicate.)
 
** check http://noc.wikimedia.org/conf/highlight.php?file=db.php
 
** check http://noc.wikimedia.org/conf/highlight.php?file=db.php
* if apache, ensure it is not an external store database
+
* If apache, ensure it is not a memcached server
** check http://noc.wikimedia.org/conf/highlight.php?file=db.php
+
* If apache, ensure it is not a memcached server or external store server.
+
 
** check http://noc.wikimedia.org/conf/highlight.php?file=mc.php
 
** check http://noc.wikimedia.org/conf/highlight.php?file=mc.php
 
* Remove from LVS pybal configuration(s).
 
* Remove from LVS pybal configuration(s).

Revision as of 12:56, 29 November 2011

Currently the process for decommissioning servers is under a state of flux. Things are being updated to make this process more standardized and streamlined.

  • Check/Update/Remove server from Server_roles : This is becoming very outdated, but should be updated and checked on servers removed.
  • If database server, ensure it is not online in any cluster and that mysql services are not running (or slaved to another server, or providing the master server for others to replicate.)
  • If apache, ensure it is not a memcached server
  • Remove from LVS pybal configuration(s).
  • Remove from DSH node lists.
  • Add the server to the decommissioning puppet manifest
  • Remove the specific node entry for the server, so only "standard" gets run.
  • Update puppet on the target server.
  • Optionally, run on sockpuppet (but Puppet will do this automatically in the following night):
# puppetstoredconfigclean.rb server fqdn
  • Remove the certificate for the server off sockpuppet (puppetca -c $fqdn).
  • Wipe data off server (To wipe disks, systems must be removed from the network. Enter an RT ticket for the onsite technician to do so in the relevant RT queue).
  • Move the server to decomissioned racks in racktables.
Personal tools
Namespaces

Variants
Actions
Navigation
Ops documentation
Wiki
Toolbox