Pmtpa cluster

From Wikitech
(Redirected from Pmtpa)
Jump to: navigation, search

The Pmtpa cluster (or Sdtpa cluster) was located at Hostway (formerly PowerMedium) and Equinix (formerly Switch and Data) data centers in Tampa, Florida (United States).


"pm" comes from PowerMedium. "TPA" is the airport code of Tampa International Airport.


  • pmtpa: Traditional services.
  • sdtpa: Networking for pmtpa.

Servers and services

Core Services/Servers - already wholly handled by eqiad - hot standby

  • The services/servers listed here are part of the primary cluster, and already have servers in EQIAD doing their job.

cp 1-10

  • cp1 1H IN A
  • cp2 1H IN A
  • cp3 1H IN A
  • cp4 1H IN A
  • cp5 1H IN A
  • cp6 1H IN A
  • cp7 1H IN A
  • cp8 1H IN A
  • cp9 1H IN A
  • cp10 1H IN A

These are sq servers. cp1 = sq51, cp2 = sq52, ...cp10 = sq60.

sq 33-106

sq33 1H IN A


sq106 1H IN A

ssl 1-4

  • ssl1 1H IN A
  • 1H IN AAAA 2620:0:860:1:208:80:152:16
  • ssl2 1H IN A
  • 1H IN AAAA 2620:0:860:1:208:80:152:17
  • ssl3 1H IN A
  • 1H IN AAAA 2620:0:860:1:208:80:152:18
  • ssl4 1H IN A
  • 1H IN AAAA 2620:0:860:1:208:80:152:19

osm-web 1-4

templates/wmnet:osm-web1 1H IN A templates/wmnet:osm-web2 1H IN A templates/wmnet:osm-web3 1H IN A templates/wmnet:osm-web4 1H IN A

Misc. Services Pending Migration

  • mobile::vumi RT #5147
  • bugzilla RT #4783 (Daniel) - pending bugzilla 4.4 deploy to zirconium. done. dzahn (talk) 03:51, 13 February 2014 (UTC)
  • observium RT #6171
  • mail/imap RT #1804
  • rancid/torrus/misc/? (streber) RT #2186 (networking) - streber services have already moved.
  • etherpad-old RT #4644 ?
  • irc-server RT #4784 - pending host in eqiad
  • "hume" RT #4792
  • APT repo/install RT #6133
  • trebuchet (git-deploy/sartoris) test host RT #6134 - can shutdown.
  • udp2log (emery,locke) RT #6143 RT #6168 (analytics) - emery still in use. locke can shutdown.
  • stat1 RT #6144 (analytics)
  • fenari (apache-deploy,,bast/homes,pybal,..) RT #6145
  • PDF service (sprint going on) RT #6149 RT #6335 pending code deploy to eqiad.
  • toolserver db replicas and jira RT #6155
  • labs RT #6158 (Ryan?)
  • dumps/dataset2, media mirror RT #6159 RT #6169 (Ariel?)
  • LDAP (OIT) RT #6163 (
  • fundraising (grosley,db78,pappas (Jeff?)
  • backups (tridge) RT #6170
  • puppetmaster (sockpuppet) RT #6175 (Alex?)
  • db9 (depends on some services) RT #6252
  • tmh1-2 (timed media/video) RT #6222


See RT #4705
See RT #4629
  • brewster 1H IN A
  • brewster is a Wikimedia APT repository (misc::apt-repository).
  • brewster is a Wikimedia Caching proxy server (misc::caching-proxy).
  • brewster is a Wikimedia haproxy host (misc::haproxy).
  • brewster is a Wikimedia Install server (misc::install-server).
  • brewster is a Wikimedia TFTP server (misc::tftp-server).
  • brewster is a Wikimedia Public Ubuntu mirror (misc::ubuntu-mirror).
  • Carbon is the replacement server for this in eqiad, it already runs the tftp server.


See RT #6252

db9 still has databases on it that need to be migrated to db1001. db9 needs decom


See RT #4783
See RT #4784
  • ekrem 1H IN A
  • ekrem is a Wikimedia IRC server (misc::irc-server).
  • ekrem is a Wikimedia MediaWiki RC to IRC relay (misc::mediawiki-irc-relay).
  • RT-4783, RT-4784 (deploy lanthanum as replacement irc server for ekrem)


See RT #4634
See RT #2046
  • emery 1H IN A
  • emery is a Wikimedia log collector (role::logging).
  • emery is a Wikimedia udp2log data collection server (udp2log::logger).
  • this will be upgraded and moved in coordination with Analytics team
  • production udp2log services have been moved off of emery (many thanks to matanya).
  • udp2log is still running on emery, would like to leave it there until we move everything to kafkatee, if possible.
    • however, if urgent, we can decom emery now.


See RT #4644
  • hooper 1H IN A
  • hooper is a Wikimedia Etherpad server (misc::etherpad).
  • * etherpad or etherpad lite in eqiad? (I suggest former, migrate service, then worry about etherpad versus etherpad lite.) --robh (talk)
  • racktables migrated to magnesium --robh (talk)
  • not etherpad-lite, that's zirconium. removed racktables from motd. what about the blog remnants? i know real blog is holmium. dzahn (talk) 00:30, 8 November 2013 (UTC)
    • just is left on it now.dzahn (talk) 00:34, 8 November 2013 (UTC)


See RT #4792
  • terbium is the eqiad replacement host.
  • hume 1H IN A
  • IPv6: 1H IN AAAA 2620:0:860:2:21d:9ff:fe33:f235
  • hume is a Wikimedia Misc - Maintenance Server: foundationwiki (misc::maintenance::foundationwiki).
  • hume is a Wikimedia Misc - Maintenance Server: pagetriage extension (misc::maintenance::pagetriage).
  • hume is a Wikimedia Misc - Maintenance Server: parser cache purging (misc::maintenance::parsercachepurging).
  • Most services have been migrated, only the below are pending migration:
  • misc::maintenance::geodata
  • misc::maintenance::update_special_pages
  • misc::maintenance::update_flaggedrev_stats
  • misc::maintenance::refreshlinks


  • kaulen 1H IN A

:* but need to clean up DNS remnants:

  • 1H IN A
  • 1H IN CNAME kaulen


See RT #5148
  • manutius 1H IN A
  • manutius is a Wikimedia Ganglia gmetad aggregator (ganglia::collector).
  • manutius is a Wikimedia central Ganglia aggregator (ganglia::monitor::aggregator).
  • manutius is a Wikimedia Torrus (misc::torrus).

pdf 1-3

  • pdf1 1H IN A
  • pdf2 1H IN A
  • pdf3 1H IN A


See RT #2186
  • streber 1H IN A
  • streber is a Wikimedia Ganglia gmetad aggregator (ganglia::collector).
  • streber is a Wikimedia network monitoring server (misc).
  • streber is a Wikimedia Really Awful Notorious CIsco config Differ (sp) (misc::rancid).
  • streber is a Wikimedia central syslog server (network) (misc::syslog-server).
  • streber is a Wikimedia Torrus (misc::torrus).
  • streber has on Apache: just "observium" and nothing on lighttpd
  • Observium, Smokeping: TODO


  • yvon 1H IN A
  • yvon is a Wikimedia IPv6 Proxy (ipv6proxy).
  • yvon is a Wikimedia SSL Proxy (sslproxy).


See RT #5147
  • zhen 1H IN A
  • This is our mobile::vumi server, paired with silver. Details in ticket.

DNS Project

  • Tampa based DNS servers are being replaced with new DNS setup. The hardware for this has already been allocated, and the below server(s) in Tampa will simply be decommissioned when it launches.
See RT #1778
  • dobson 1H IN A
  • dobson is pmtpa's primary DNS resolver
  • dobson is a Wikimedia authoritative DNS server (wikimedia-task-dns-auth).
See RT #437
  • linne 1H IN A
  • linne is ns1
  • linne is also url-downloader 1H IN A  ; linne

Labs Infrastructure

  • Labs is being specifically handled by Ryan Lane.
  • wikitech 1H IN A
  • frlabs 1H IN A
  • labs-ns0 1H IN A
  • virt0 1H IN A
  • Ryan Lane will handle this, do not move it. --robh (talk)

Misc. Services Migrated

dataset 2

  • All dumps run on dataset1001. A few misc things need to be moved, and then the last thing will be the web server and some cnames.


  • capella 1H IN A
  • capella is a Wikimedia IPv6 tunnel relay (6to4/Teredo) (role::ipv6relay).
  • Consider this network equipment. We can afford downtimes there too. faidon (talk)
  • Server nitrogen handles these exact same puppet classes. --robh (talk)


  • was blog server, migrated to magnesium --robh (talk) 22:34, 10 May 2013 (UTC)
  • System is now in sandbox vlan for investigation of the intrusion.
  • Once that is done, system should have ticket entered to be wiped and have network ip reclaimed.
  • It's still in DNS. dzahn (talk) 23:36, 14 May 2013 (UTC)



See RT #4781
See RT #5018
See RT #5188
See RT #5258
  • is left. dzahn (talk) 19:51, 7 May 2013 (UTC) . Done. moved to zirconium. dzahn (talk) 18:45, 7 June 2013 (UTC)
  • Moved or removed services/sites: secure.wm moved to primary cluster, retired communicate.wo, outreachcivi.wo, ocs.wikimania2009.wo, wm08reg.wo, wm09chols.wo, wm10schols.wo
  • Has been moved from singer to cluster. Is now an Apache config "secure.wikimedia.conf" in git running on cluster. dzahn (talk) 19:50, 7 May 2013 (UTC)
singer has been completely removed from DNS and puppet today. dzahn (talk) 18:45, 7 June 2013 (UTC)


See RT #3645
  • OTRS Server, has been migrated to iodine
  • backups of williams:/opt & williams:/etc are at iodine:/root/*tgz
  • is now a CNAME of

Misc. Services Abandoned


See RT #3213
See RT #5155

alsted 1H IN A removed from DNS dzahn (talk) 23:23, 10 May 2013 (UTC)

  • decommission and wipe host, no longer needed, used to be swift test host. --robh (talk) 18:03, 15 May 2013 (UTC)

bellin / blondel

See RT #5120
  • bellin & blondel are m1 shards, which have been replicated by eqiad servers already --robh (talk) 22:11, 10 May 2013 (UTC)


See RT #5154
  • Had nothing running, was spare server, production dns removed and mgmt left in-tact; will be wiped in preparation of shipment.


  • This is Martin's OTRS sandbox host. This will NOT migrate, and will simply stop being used when he is done rolling upgrades. Do NOT remove from DNS. --robh (talk) 18:14, 14 May 2013 (UTC)
  • db61 1H IN A

ocg 1-3

See RT #1593
See RT #5154
  • DNS Entries removed for production, mgmt updated to just asset tag, servers being wiped. --robh (talk)

owa 1-3

See RT #2511
See RT #5143
  • owa1 1H IN A
  • owa2 1H IN A
  • owa3 1H IN A
  • has address
  • removed from DNS, replaced mgmt entries with Dell service tags.
  • <@mutante> +WMF3848
  • <@mutante> +WMF3847
  • <@mutante> +WMF3846

dzahn (talk) 22:35, 14 May 2013 (UTC)

  • 1H IN A This is the laptop IP for our on site tech, leave in place. --robh (talk) 22:42, 14 May 2013 (UTC)

Old DNS Cleanup

  • These servers were decommissioned before the current review of all servers, and have just been removed from DNS and any other cruft left over in configuration files.


  • albert? ldap 1H IN A  ; albert - remove from DNS .. not used
  • dzahn: remove, points to which is
  • albert and that was an old silicone mechanics box purchased in 2004
  • removed from DNS - dzahn (talk) 23:23, 10 May 2013 (UTC)


  • anthony 1H IN A
  • "? this ends up on pdf3. eh ?"
  • removed from DNS. dzahn (talk) 19:16, 1 May 2013 (UTC)


  • bayle 1H IN A
  • this has been decommissioned
  • removed from DNS dzahn (talk) 23:37, 10 May 2013 (UTC)


  • browne 1H IN A
  • dzahn - remove host browne (decom'ed, gone). removed from DNS dzahn (talk) 23:21, 10 May 2013 (UTC)

dataset 1

  • was long gone, killed dns entries --robh (talk) 17:57, 15 May 2013 (UTC)


  • gilman 1H IN A
  • --> root@pdf3  ?
  • < cmjohnson1> mutante: khaldun (gone...bought in 2004 silicone mechanics)
  • gilman has been decommissioned but we still have it on the floor in tampa
  • removed from DNS dzahn (talk) 23:38, 10 May 2013 (UTC)


gurvin 1H IN A

gurvin is a Wikimedia ISSLPv6 Proxy (ipv6proxy).

gurvin is a Wikimedia SSL Proxy (sslproxy).

This wasn't true anymore, the server was unused and has been shutdown/decom. on the way to being wiped. dzahn (talk) 01:40, 15 November 2013 (UTC)


  • khaldun 1H IN A
  • < cmjohnson1> mutante: khaldun (gone...bought in 2004 silicone mechanics)
  • removed from DNS dzahn (talk) 23:39, 10 May 2013 (UTC)


  • msfe-pmtpa-test 1H IN A ;owa1
  • 1H IN A ;owa2
  • 1H IN A ;owa3
  • also see owa 1-3, old swift testing boxes, about to be decom'ed
  • alias for owa1
  • already removed from DNS dzahn (talk) 23:06, 14 May 2013 (UTC)


  • project2 1H IN A
  • removed from DNS. dzahn (talk) 19:17, 1 May 2013 (UTC)


  • pdf4 - pdf15 : these never existed, killing dns entries--robh (talk) 22:38, 14 May 2013 (UTC)


  • yongle 1H IN A
  • this is actually also "ekrem"
  • removed from DNS. dzahn (talk) 19:17, 1 May 2013 (UTC)

Fundraising Infrastructure - Jeff Green

  • Servers used in fundraising are behind a bit of security, so Jeff will have to review and confirm all the fundraising allocated machines in Tampa are migrated.


See RT #5135

db78 db78.pmtpa.wmnet has address

Runs precise. Offsite fundraising database and log/db/data archive. Do not kill.


See RT #1697
See RT #3022
  • grosley 1H IN A
  • Jeff: " Runs lucid. Warm standby for aluminium, but does not host critical data/services. Do not kill this host without deploying a warm standby at eqiad."
  • grosley is a Wikimedia fundraising sites and operations (misc::fundraising).
  • grosley is a Wikimedia fundraising backup archive (misc::fundraising::backup::archive).
  • grosley is a Wikimedia fundraising impression/banner log archive (misc::fundraising::impressionlog::archive).
  • grosley is a Wikimedia fundraising mail server (misc::fundraising::mail).
  • grosley is a Wikimedia jenkins integration server (misc::jenkins).
  • < pgehres> it should be kept around as a hot failover for fundraising
  • < pgehres> but it is only a backup for aluminium
  • < pgehres> if tampa died, we would lose nothing
  • < pgehres> just redundancy


pappas 1H IN A

Jeff: "Runs precise. Redundant bastion server for fundraising."

payments 1-4

payments1 1H IN A

payments2 1H IN A

payments3 1H IN A

payments4 1H IN A

Payments1 is in production as the master db for payments-wiki, payments1001 can take over easily but runs a newer mysql rev and can't be master for payments1-4.

Payments2-4 are idle. All are available as hot-standby.

Unsorted Services

We want to move all of these out of Tampa.



See RT #2350

fenari 1H IN A

bastion, deployment host (separate deployment from bastion role, use bast1001 and fluorine (--> RT-2350)

Also hosts; users' ~/public_html is served as, and this has been used in the past for links that should persist


See RT #4634
See RT #4720

locke 1H IN A

locke is a Wikimedia log collector (misc::log-collector).

locke is a Wikimedia udp2log data collection server (misc::mediawiki-logger).

locke is a Wikimedia log collector (role::logging).

locke is a Wikimedia udp2log data collection server (udp2log::logger).


See RT #1804

mchenry 1H IN A

+ IPv6 1H IN AAAA 2620:0:860:2:219:b9ff:fedd:c027

mchenry is Wikimedia's primary mail relay, and pmtpa's secondary DNS resolver.

mail 1H IN A  ; service IP on mchenry

wiki-mail 1H IN A  ; service IP on mchenry


ms10 1H IN A

ms10 is a Wikimedia Media mirror (rsync access for external mirrors) (role::mirror::media). dzahn (talk) 23:04, 14 May 2013 (UTC)


raskin.usability 1H IN A


See RT #4346

sanger 1H IN A

smtp, imap


See RT #2183
See RT #3066

spence 1H IN A


See RT #6144

stat1 1H IN A

stat1 - analytics host

storage 1-2

storage1 1H IN A

storage2 1H IN A

1H IN AAAA 2620:0:860:2:230:48ff:fe5a:eb1e

deleted from DNS. dzahn (talk) 21:56, 17 May 2013 (UTC)


tridge 1H IN A ; Cold Backup Server

backup server TODO!

yea, these are backups and they are still in Tampa


Media storage (Swift)



ns0/ns1 - bayle and linne . well , obviously DNS


recursor0 1H IN A  ; bayle recursor1 1H IN A  ; mchenry


observium 1H IN A

cmjohnson wanted to take a look at moving this one

retired services

-- pre-labs things --

-- misc-- (this was already commented out in the past) (we got rid of the Linode instance)

-- kept for backwards compatibiliy --    (but if your client uses that, please update to "")

-- kept for historic reasons --

long live .. :)

--  backed up docroots from singer on tridge --

blog.tar.gz            bugzilla.tar.gz  outreachcivi.tar.gz  svn_blog.tar.gz      techblog.tar.gz       wm09schols_info.txt.gz  wm10reg.tar.gz
bugzilla.older.tar.gz  bzapi.tar.gz     survey.tar.gz        svn_techblog.tar.gz  wikimania2009.tar.gz  wm09schols.tar.gz       wm10schols.tar.gz  .. and more

-- about to be moved --


See also