On Apr 21, 2009, at 1:21 PM, PS SP wrote: > I am considering SQLite for a project that would create a fairly large > database (order of 50 million rows, max 100 million rows). I'll be > using the > C API. The DB is mostly used for queries with relatively fewer > inserts and > updates. All updates and inserts will be done in batches and will be > scheduled once or twice per day. > While discussing this with other folks at work who have used SQLite > in the > past, I heard someone say that they have experienced sporadic database > corruptions in the field. Unfortunately more details are not > available at > this point. The only other information is that the environment is > Windows. > This is obviously a source of concern for me. > > Has anyone else experienced DB corruptions? Are there any things > that could > trigger DB corruption -- for example, updates to indices on large > data sets? > For example, in my use case because I can do batch updates, dropping > indices > before updates and inserts and recreating them later is an option > (if it > removes the possibility of DB corruption). Any input on this would > be very > valuable to me and the community.
Background material on this is at http://www.sqlite.org/atomiccommit.html and especially section 9.0: Things that can go wrong. There have been bugs in SQLite in the past that would result in database corruption. But they have been very obscure and they are now all fixed as far as we know. And we do test for this kind of thing very carefully. http://www.sqlite.org/testing.html Most database corruption problems come about when hardware loses power in the middle of a transaction and the disk synchronization primitives in the operating system or disk controller or disk hardware do not work as advertised. D. Richard Hipp d...@hwaci.com _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
