We concurrently insert into the database from 4 clients, we query from another 3 clients and we do some misc. things from another client (e.g. alter table add column, create additional index, drop/add index).
We can calculate some invariants to check in the end, in the regression test. e.g. number of total rows in the table, result of a query after all this load is done.
We can achieve this with pg_regress. To do that we just should avoid creating a separate DB for each test and should instead use pg_regress like most people usually use it.
I am open to keeping current tests intact and just adding a new mechanism for running this kinds of tests. I am open to other suggestions as well.
We should have stress tests such as:
We concurrently insert into the database from 4 clients, we query from another 3 clients and we do some misc. things from another client (e.g. alter table add column, create additional index, drop/add index).
We can calculate some invariants to check in the end, in the regression test. e.g. number of total rows in the table, result of a query after all this load is done.
We can achieve this with
pg_regress. To do that we just should avoid creating a separate DB for each test and should instead usepg_regresslike most people usually use it.I am open to keeping current tests intact and just adding a new mechanism for running this kinds of tests. I am open to other suggestions as well.