On Tue, Feb 16, 2021 at 01:49:02PM +0000, Guttman, Maoz wrote: > Hi, > > Problem statement: > I have to develop a solution in which a single source populates a table. Once > the table is populated, it is considered as read-only and then we run many > read-only queries on it. > Such read-only tables are generated by multiple simulation runs: each > simulation populates an independent table, meaning there is no cross-write to > the tables.
I don't think you can share data dir or its files across clusters. Things like autovacuum will want to be able to write to the data dir. I don't think you can create a table-am which handles only "read-only" operations - currently they're all required. Did you think about using FDWs for this, instead ? Otherwise, maybe you could make a copy of the cluster (like with a filesystem clone) or pg-basebackup, or you could use replication. -- Justin
