The last database dump on download.wikimedia.org is from March 9th, almost one month ago.
Is it correct therefore that we have no backup for data written after that day? Or is there some other backup strategy at the colo?
Alfio
Alfio Puglisi wrote:
The last database dump on download.wikimedia.org is from March 9th, almost one month ago.
Is it correct therefore that we have no backup for data written after that day? Or is there some other backup strategy at the colo?
Alfio
We have several (three?) databases receiving content replication. So even if 2 servers totally crash at the same time, and 1 can not be used for whatever reason, we will still have 1 working copy.
ASCII schema is something like:
--> Slave #1 / Users -> Master Database -----> Slave #2 \ --> Slave #3
As for regular dumps (aka twice per week), stats generation ... I think we wait for a monster server (read more disk space).
cheers,
Ashar Voultoiz wrote:
Alfio Puglisi wrote:
The last database dump on download.wikimedia.org is from March 9th, almost one month ago.
Is it correct therefore that we have no backup for data written after that day? Or is there some other backup strategy at the colo?
Alfio
We have several (three?) databases receiving content replication. So even if 2 servers totally crash at the same time, and 1 can not be used for whatever reason, we will still have 1 working copy.
ASCII schema is something like:
--> Slave #1 /Users -> Master Database -----> Slave #2 \ --> Slave #3
As for regular dumps (aka twice per week), stats generation ... I think we wait for a monster server (read more disk space).
cheers,
Hoi, Yes we have slaves and it is NOT a back-up strategy. The last big database outage proved that point, we were saved by the fact that one slave was out of action. This allowed us to restore functionality. It is good to have slaves but when the database is corrupted you end up with nothing. Having slaves is not an alternative for a backup. Thanks, GerardM
wikitech-l@lists.wikimedia.org