[Toolserver-l] Changes to replicated wiki databases

Hi, Over the next couple of weeks we will be making some changes to the way that wiki databases work on the Toolserver. Currently, each cluster (s1..s6) has only a single Toolserver slave. (For example, s1 is on yarrow and no other server). We will be adding an additional server to each cluster, so each cluster has two servers. The existing server will be designated the 'standard' server, and the new server will be a 'fast' server. The standard server can be used as it is now, with no changes. The 'fast' server will be the same as the standard server, with two exceptions: * creating user databases will not be allowed * queries running longer than 60 seconds will be automatically killed The intent here is that by preventing long running queries on the fast server, replication lag will be much less likely, and queries that need the most up-to-date data, which are usually also fast queries, will be able to use the fast server. If either server fails, its queries will be redirected to the other server. So, if the s1 fast server fails, all fast queries will go to the standard server. If the standard server fails, all queries will go to the fast server. If you want your queries to run on the standard server, you do not need to do anything. The current server for each cluster, where the existing 'sql-sX' aliases point, will become the standard server. if you want your queries to run on the fast server, you should connect to 'sql-sX-fast' instead of 'sql-sX'. You cannot connect to the fast server using DNS aliases; you must use the sql-sX aliases. Currently, no fast servers are available, so all sql-sX-fast aliases point to the standard server. The first cluster to get a fast server will be s1, which I hope will happen next week. One cluster will not get its fast server until all projects have finished migrating from the stable server, as the stable server host will be converted into the 3rd fast server. - river.