Martin, Thanks :) >Running vacuum is the right solution, but I think you have to let it >finish. In particular, in that version a database-wide vacuum has to >complete before it will update the datfrozenxid (it's not tracked per >table). >> a) is my assumption about the database being ok correct - assuming all >> tables have been vacuumed recently, including catalog tables? >Should be ok, but apparently you missed one, or didn't do a database wide >vacuum. Yes, probably missed this 220 million row beast that's still running now.. If I was to abort this vacuum, given that all other tables are vacuumed (including system catalog tables), what's the worst case scenario? - given that more transactions are happening on the database If I understand correctly, it would be that some rows could disappear from this large unvacuumed table if their xid was too old - but no other consequence? (fully aware that a db-wide vacuum is needed, but if it can [safely] wait for the weekend that would be preferable) Many thanks, S -----Original Message----- From: Martijn van Oosterhout [mailto:kleptog@xxxxxxxxx] Sent: 24 October 2006 10:24 To: Shane Wright Cc: pgsql-general@xxxxxxxxxxxxxx Subject: Re: [GENERAL] recovery from xid wraparound On Tue, Oct 24, 2006 at 07:43:15AM +0100, Shane Wright wrote: > Anyway - not noticed any data loss yet and was hoping it would be such > that if all tables had been vacuumed recently (including system > catalog tables), that there would be no remaining rows that would > appear to have a future xid and so the database should be ok? Running vacuum is the right solution, but I think you have to let it finish. In particular, in that version a database-wide vacuum has to complete before it will update the datfrozenxid (it's not tracked per table). > a) is my assumption about the database being ok correct - assuming all > tables have been vacuumed recently, including catalog tables? Should be ok, but apparently you missed one, or didn't do a database wide vacuum. > b) is it possible to safely abort my whole table vacuum now so I can > run it at the weekend when there's less traffic? Aborting vacuum is safe, but you have to do a database-wide vacuum at some point. > c) if I have experienced data loss, on the assumption all the table > structure remains (looks like it does), and I have a working backup > from before the xid wraparound (I do), can I just reinsert any > detected-missing data at the application level without needing a > dump/reload? A VACUUM will recover any data that slipped beyond the horizon less than 1 billion transactions ago, which I think covers you completely. The only issue is that unique indexes may be confused because new conflicting data may have been inserted while the old data was invisible. Only you can say if that's going to be an issue. Hope this helps, -- Martijn van Oosterhout <kleptog@xxxxxxxxx> http://svana.org/kleptog/ > From each according to his ability. To each according to his ability > to litigate.
