Remove flaky negative walrep test.
This scenario intended to be tested is if primary (walsender) exits, walreceiver must exit as well. Then wal receiber should come-up and try to reconnect and fail again till primary is not brought back. Once primary is up, it should be able to connect. The way is coded today is extremely hacky logic as simulates via fault injection by creating file `wal_rcv_test` and providing option to suspend at what point. Then signal.SIGUSR2 is sent to notify standby to inject the fault. There exists no mechanism to check if fault was hit or not. So, tests tend to little unreliable at times since don't know if fault was hit or not, hence large sleeps are used in this test. Plus, also file wal_rcv.pid is created in code just for testing purpose to validate some behaviors. Hence, removing this flaky time-consuming test. (cherry picked from commit ff9c80cc)
Showing
想要评论请 注册 或 登录