Currently we are using READ write pattern with transaction wrapper to see if version was changed when checking for conflict. This pattern may not be valid - in fact when transaction is set there is no locking for column.
This means that updates still can happen. We need to investigate other options. Happy for us there are tons of really good materials online on this problem (this is really well understood problem in database word) so we may just investigate possible workarounds and present them to team.
See: https://blog.2ndquadrant.com/postgresql-anti-patterns-read-modify-write-cycles/ |
|
