The overall load characteristics of the app are that there are many readers, doing queries that are *usually* sub-second, and occasional large batch inserts/deletes of, say, hundreds of thousands of rows. The main goals are to not block readers, and minimize exposure of readers to incompletely loaded batches. So, if a transaction per batch is too aggressive, or blocks readers, then we need to implement "not finished loading" flags on various tables that the readers can use to filter out the half-loaded batches.
On Tuesday, May 27, 2014 10:45:29 AM UTC-7, Brian Craft wrote: > > Version 1.3.171. The only option I'm setting is MVCC, like > "somefile;MVCC=TRUE". I'll try to get a thread dump in a few minutes. > > Does this mean readers should not be blocked during the insert? > > On Tuesday, May 27, 2014 4:18:01 AM UTC-7, Noel Grandin wrote: >> >> Can you post a thread dump, so we can see where it is blocked? >> Also which version is this, and what does your DB URL look like? >> >> On 2014-05-25 20:24, Brian Craft wrote: >> > I'm doing a large "insert into .. select * from CSVREAD ...", which is >> in a transaction. It seems to be blocking >> > readers, regardless of whether MVCC is enabled. Is this expected? If I >> understand the threads on MVCC, it will allow >> > -- You received this message because you are subscribed to the Google Groups "H2 Database" group. To unsubscribe from this group and stop receiving emails from it, send an email to h2-database+unsubscr...@googlegroups.com. To post to this group, send email to h2-database@googlegroups.com. Visit this group at http://groups.google.com/group/h2-database. For more options, visit https://groups.google.com/d/optout.
