On 17 Jul 2018, at 8:37am, Rob Willett <rob.sql...@robertwillett.com> wrote:
> I suspect that part of the issue is the VPS provider we use has a rate > limiter on IOPS which is not normally an issue for us, but that might have > slowed it down somewhat. However I don't think that it would have slowed it > down by hours. Actually I think VPS had a lot to do with the time the operation took. Any kind of virtual machine takes a terrible hit during the sort of storage access involved in dropping the table. SQLite spent that whole time accessing your 50GB database file in an apparently random order. So you had nine hours of cache misses, causing the virtual machine to continually write virtual pages back to real storage and read other pages into memory. Virtual systems are optimized for cache hits, not cache misses. I can't prove it without a lot of pointless data manipulation on your type of VPS, but I think you found its least optimal operation. The good part is that now your database is less than 1GB long you're going to see a massive increase in speed since the whole database may well fit in the cache of your virtual machine. Must remember in future, when people report unusually slow operations, to ask whether they're using a virtual machine or real hardware. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users