Exactly, essentially it could be as simple as whoever was making the database changes making a slip and changing the wrong field or something similar. Again though it seems to come down to good practice, why were such devestating changes made to a live server? When Mythic were testing clustering they brought up a new test server - Hector to cluster it with Pendragon, the new server was an exact copy of one of the live servers, I beleive it was Palomides or Percival.Briannon wrote:If this is true then I am amazed at this. Mythic clustered 21 servers when they first clustered, and they have since made further clustering changes to make some 4-server clusters. In all that work none of the server databases were corrupted (as far as I know) and there certainly hasn't been any down time beyond the few couple of hours for patches and changes.
If database changes for clustering just two servers is one of the reasons for this problem then why has it only happened on the euro servers this particular time, and not on any of the clustering done on the US servers?
It would surely make sense to make a copy of one of the servers to emulate clustering with Gorre or another copied server and then if it works copy the fully tested and working version over instead of breaking the live server from the off. Did GOA do this? Emulate the changes internally on a copy of a live server? If so at least we can rule out lack of a proper system of testing - something which has plagued GOA heavily in the past.
Really just don't see how Mythic can be blamed when their servers run so flawlessly in comparison.