I've just now subscribed to this list, so please stop me if this has been already discussed... After hearing about the storm coming toward the wikimedia servers. I thought to myself, wouldn't it be a good idea to have backup servers elsewhere in the world? Especially since Florida is highly susceptible to this sort of weather this time of year... I'm not sure if our funding or budget would allow for distribution of our servers. However, my guess is, there are probably more than one wikipedia people who wouldn't mind setting up a server of theirs somewhere as a backup. Even if it's just a no edit cached back-up. Something is better than nothing...
-- Michael Becker
mbecker wrote:
I've just now subscribed to this list, so please stop me if this has been already discussed... After hearing about the storm coming toward the wikimedia servers. I thought to myself, wouldn't it be a good idea to have backup servers elsewhere in the world?
I think it would be a good idea, but only if the expense made sense.
Here are three things I wanted to pass along:0
1. Keep in mind that our hosting facility remained confident throughout the ordeal that there would be no loss of connectivity due to this storm. It's a strong building, they have good power contingency plans, etc. We are not in a position to evaluate their confidence, of course. I merely pass it along as information...
It's pretty hard to take down a data center like this one.
2. We will soon have functional squids in France. As we move forward with that, getting them tested, gaining confidence in our ability to manage that, then we should probably consider putting a db/backup server there, or something similar.
3. We are widely beloved, and it is not difficult for us to get very fast offers of hosting on a temporary basis. In this particular case, the wonderful people at Oregon State University's Open Source Labs offered us "3-4 hefty machines" available at a moment's notice. I am going to talk to them some more, but it seems likely that if we pay the cost of a machine, they would be happy to host on a permanent basis a slave db machine which we could keep synced to have a constant realtime backup.
--Jimbo
On Mon, 16 Aug 2004 09:03:24 -0700, Jimmy (Jimbo) Wales jwales@wikia.com wrote:
- We are widely beloved ... it seems likely that if we pay
the cost of a machine, they would be happy to host on a permanent basis a slave db machine which we could keep synced to have a constant realtime backup.
At some point soon, we will be backing up many terabytes of archives and historical data. Would it make sense to have an offsite location dedicated to preserving these backups and regular snapshots of the db, in case there is ever data corruption which goes unnoticed and is replicated to all active slaves/mirrors?
That location could also maintain a synced mirror of the current projects and key filesystem content, along with a few machines available for configuration as temporary web and db servers, in case something terrible happened at the colo facility.
+sj+
Yes, all the things you say "make sense". :-) It's really just a matter of priorities. I remember making negative comments about such suggestions when we had only 2 or 3 machines, basically arguing that worrying about destruction of the colo was pointless when we could hardly function with the loss of a single *machine*.
But with 14 (15?) machines in service now, and with 10 more on the way, with 12 4U machines soon to arrive as well (though not to be pressed into service right away until we figure out what to do with them exactly), and with the soon-to-exist Paris data center, it does start to make some sense to think about geographical diversity.
I'm not sure how important this is, really. But it is at least worth thinking about.
Sj wrote:
On Mon, 16 Aug 2004 09:03:24 -0700, Jimmy (Jimbo) Wales jwales@wikia.com wrote:
- We are widely beloved ... it seems likely that if we pay
the cost of a machine, they would be happy to host on a permanent basis a slave db machine which we could keep synced to have a constant realtime backup.
At some point soon, we will be backing up many terabytes of archives and historical data. Would it make sense to have an offsite location dedicated to preserving these backups and regular snapshots of the db, in case there is ever data corruption which goes unnoticed and is replicated to all active slaves/mirrors?
That location could also maintain a synced mirror of the current projects and key filesystem content, along with a few machines available for configuration as temporary web and db servers, in case something terrible happened at the colo facility.
+sj+ _______________________________________________ Wikitech-l mailing list Wikitech-l@wikimedia.org http://mail.wikipedia.org/mailman/listinfo/wikitech-l
wikitech-l@lists.wikimedia.org