On 11/03/2014 05:27 PM, Sean Pringle wrote: > 1. We need some memory and time limits for user queries. Memory usage is > easy to track server-side on a per-client basis, but users may find it > difficult to predict or understand why specific queries trip some > arbitrary memory limit. So, just time based? Thoughts?
I'd go for memory despite the fact that it's a bit more obscure to the enduser because that's the actual proximate cause of the failures - time based would almost certainly catch /almost/ all of the problematic queries but it's hard to predict which ones would baloon faster. > 2. The TokuDB bug DB-739 appears only on specific types of upstream > transaction, so some replica tables (including but not necessarily > limited to *links, user, recentchanges, and geo_tags) are being switched > back to InnoDB until further notice. The bug is stated to be fixed upstream, do we know how long before that filters down into a version we can deploy? > > 3. After #2 we resync across the board, yet again. Either way. -- Marc _______________________________________________ Labs-l mailing list [email protected] https://lists.wikimedia.org/mailman/listinfo/labs-l
