I want to change DB schema for "cu_changes" table for implement Global CheckUser.
but I guess when merge the each clusters "cu_changes" to one cluster, single DB cluster IO is expensive.
WMF DBAs are how to handling that? horizontal split? or should be still holding splited write for the each DB clusters?
- Plan 1: Just simply add "cuc_wiki" field to "cu_changes" and merge the table to single cluster
- Plan 2: Just simply add "cuc_wiki" field to "cu_changes" and Create master table "sN_cu_changes" and merge the each wiki's table to the master table. (I think this is same IO volume per each clusters); N is cluster number. e.g. s3_cu_changes
- Plan3 : Holding current DB schemas. When issued a query, just SELECTing each DBs (I guess it is too slow for getting result)