I think we can fix CAY-2040 by ensuring a separate transaction (maybe in a separate thread) for filling the app PK cache. The more I think about it, the more sense it makes. It is really separate from the main tx.
Andrus > On Dec 4, 2015, at 12:45 PM, Aristedes Maniatis <[email protected]> wrote: > > On 3/12/2015 2:08pm, John Huss wrote: >> The main advantage of a native solution is that you know it works. If you >> ever used native PK generation the AUTO_PK_SUPPORT table feels very >> kludgy. It works (I guess?), but it's not ideal. A native solution is >> faster and simpler. It handles transactions better. Most DBs (all?) do PK >> generation outside of transactions, so incrementing the sequence always >> persists regardless of transaction rollbacks or commits. > > Thanks for your detailed reply. Transactions (this bug) > https://issues.apache.org/jira/browse/CAY-2040 are exactly why we want to > change. Perhaps this is not really a bug, but just a documented reason not to > use this approach and limitation that would be very hard to work around. > > > Ari > > > > -- > --------------------------> > Aristedes Maniatis > GPG fingerprint CBFB 84B4 738D 4E87 5E5C 5EFA EF6A 7D2E 3E49 102A
