UID

reserved UIDs & GIDs

Althugh we try to keep this up-to-date it the authoritative source is admin.yaml

Make sure to add reservation entry to admin.yaml

If you want the account to be created every where you must reserve the UID in the puppet admin module and use systemd-sysuser. like in this example.

If the user will just exist on a few machines then you should resever the account in the admin module with a commented block. like in this example, then create the user with a normal user block e.g.

    systemd::sysuser { 'git':
      content    => [
      'usertype' => 'u',
      'name'     => 'git',
      'id'       => 915:915,
      'gecos'    => 'git used by GitLab',
      'home_dir' => '/var/opt/gitlab',
      ]
    }

• (table columns are sortable)

UID	GID	user name
33	33	www-data
48	48	apache
107	112	puppet
110	115	nagios
111	116	mwdeploy
444	444	gerrit2
445	445	rancid
498	498	phd (phabricator)
499	499	trebuchet
901	901	reprepro
902	902	swift
903	903	hdfs (previously jenkins)
904	904	yarn
905	905	mapred
906	906	analytics
906	906	analytics
907	907	druid
908	908	hadoop
909	909	analytics-privatedata
910	910	analytics-product
911	911	analytics-search
912	912	analytics-research
913	913	analytics-platform-eng
914	914	(available for use)
915	915	git
916	916	kafka
10002	10002	l10nupdate

permission/security hierarchy

the security hierarchy looks as follows as decribed by TimStarling:

• root > wikidev > mwdeploy > www-data
  • root can own wikidev but wikidev can't own root
  • wikidev can own mwdeploy but mwdeploy can't own wikidev
  • scripts owned by mwdeploy can only be run by www-data
  • everything has to su to www-data before running maintenance scripts

also see: task T79786