Obsolete:Backup procedures
This page contains historical information. It may be outdated or unreliable.
THIS PAGE IS OLD AND OUTDATED. It is kept for historical reasons. Most of the info inside will either be or already is wrong/irrelevant. Backup infrastructure has been replaced by Bacula
This page is an inventory of our backups for both clusters data and office data.
Some definitions:
- On-site
- here means the backup is in the same physical location as the master, but on a different machine.
- Off-site
- means the backup is in a physically separate location. In most cases it's nice to have both!
- Red cells
- denote a lacking area which we want/need.
Service backup state
| Service | Master | On-site | Off-site | Notes | Approx. Space Needed |
|---|---|---|---|---|---|
Wiki primary data | |||||
| Wiki main databases | EQIAD db* | EQIAD slave replication, LVM snapshots every 8 hours (have we tested restoring from one?) | PMTPA db slaves & snapshots / ESAMS toolserver replication (how complete is the toolserver repl?) | http://noc.wikimedia.org/dbtree/ | 4 TB - check |
| External storage | EQIAD db* | EQIAD slave replication | PMTPA - ? /ESAMS toolserver replication - how complete are these? | 5 TB - check | |
| Wiki data dumps | PMTPA Obsolete:dataset2 | Gluster storage (labstore1 under /publicdata-project directory) Last 5 good only. | EQIAD: via rolling rsync Obsolete:dataset1001; via daily rsync, all public data: dumps.wikimedia.your.org, last 5 good dumps: wikipedia.c3sl.ufpr.br | See Dumps/Mirror status for more info. | 6T (last 5 good)/ 29T (all) - update these |
Images and media | |||||
| Uploads | PMTPA Swift | 3 copies per file | EQIAD Ceph | 23T? | |
| Thumbs | PMTPA Swift | 3 copies per file | EQIAD Ceph | 12T? | |
Software development and configuration | |||||
| MediaWiki config | PMTPA NFS | tridge | Subversion | ||
| Apache config | PMTPA NFS | tridge | Automatic | rsync | |
| Subversion | Formey | tridge | 3 GB | ||
| Gerrit | Manganese | Formey | Databases on db9/10 | 3 GB | |
| Bugzilla data | PMTPA db9 | db10 replication | EQIAD db1008/db1025 & snapshots | 1 GB | |
| Bugzilla config | PMTPA db9 | db10 replication | EQIAD db1008/db1025 & snapshots | db10.pmtpa.w /a/sqldata 0 66532680 9020197 13.6 31:49 4724.1 31:49 4724.1 | 1.5 GB |
| Bugzilla frontend | PMTPA isidore | daily rsync to Tridge | |||
| Wikitech wiki | linode | PMTPA tridge; ESAMS ? | |||
Communications | |||||
| OTRS data | PMTPA db9 | db10 replication | EQIAD db1008/db1025 & snapshots | 57 GB | |
| OTRS config | PMTPA bart | 19 GB | |||
| blogs frontend data | PMTPA singer | Daily backups to Tridge | |||
| bugzilla | PMTPA isidore | Daily backups to Tridge | |||
| IMAP mail | PMTPA sanger | PMTPA mchenry, Daily backups to tridge | see Mail#Backups | 32 GB | |
| Mailing lists | ESAMS lily - decommissioned | EQIAD sodium | PMTPA mchenry | See Mailing lists#Backups | 50 GB |
| DNS | EQIAD ns0, PMTPA ns1 | in gerrit | ESAMS ns2 | In gerrit | |
| Google docs | Google :P | ad-hoc downloads | |||
Soft data | |||||
| Search databases | PMTPA various | EQIAD is primary now | SDTPA | ||
| HTTP logs | PMTPA locke | emery | /a/squid | oxygen (eqiad) | |
| MediaWiki logs | PMTPA NFS | tridge | fluorine | /home/wikipedia/logs | |
Fundraising | |||||
| Fundraising Front-End | EQIAD aluminium | software:svn logs:storage3->tridge |
puppet:files/misc/scripts/offhost_backups | 5GB | |
| Fundraising Databases | EQIAD db1008 | db1025 replication + snapshots | PMTPA storage3 replication, snapshots, dumps tridge daily rsync of dumps |
puppet:files/misc/scripts/dump_fundraisingdb | 1TB (15GB/day) |
| Payment Processing | software:svn db+logs:PMTPA |
db:mysql replication db+logs:encrypt+copy daily to tridge |
db+logs:replicate to silicon | payments cluster | 100GB(?) |
| Impression Logs | PMTPA nas1 | nas1001 | 1.5TB year | ||
Misc data | |||||
| Server home directories | PMTPA nfs1 / nfs2 | tridge | Daily backups to Tridge | 30 GB | |
| stat1 /a and /home | - | Amanda to tridge | gerrit change | 400 GB | |