From Wikitech
Jump to navigation Jump to search
This page contains historical information. It is probably no longer true.

Wikimedia servers and software make heavy use of clustering to simplify management of similar hosts and services. Currently this is done in a non-standardized, ad-hoc way however, which means that e.g. Ganglia, Nagios, Puppet, dsh etc. all use different category names. Because this will become even more important with the addition of a second primary data center, this page attempts to put some thought into standardizing it all.


We can distinguish at least the following levels in our infrastructure:

(pmtpa and esams)
physical locations
(pmtpa, sdtpa, esams, knams)
subnets (public subnet, squid-lvs subnet, internal subnet)
host groups
(cache servers, database servers, nameservers)
service groups
(mediawiki application, memcached, squid, mysql)
further subdivisions
(frontend squid)

See also