On Thu, Dec 30, 2021 at 4:18 AM SATYANARAYANA NARLAPURAM <satyanarlapuram@gmail.com> wrote: > > On Wed, Dec 29, 2021 at 2:04 PM Tom Lane <tgl@sss.pgh.pa.us> wrote: >> >> SATYANARAYANA NARLAPURAM <satyanarlapuram@gmail.com> writes: >> > I noticed that below critical replication state changes are DEBUG1 level >> > logged. Any concern about changing the below two messages log level to LOG? >> >> Why? These seem like perfectly routine messages. > > > Consider a scenario where we have a primary and two sync standby (s1 and s2) where s1 is a preferred failover target and s2 is next with synchronous_standby_names = 'First 1 ('s1','s2')'. In an event, s1 streaming replication is broken and reestablished because of a planned or an unplanned event then s2 participates in the sync commits and makes sure the writes are not stalled on the primary. I would like to know the time window where s1 is not actively acknowledging the commits and the writes are dependent on s2. Also if the service layer decides to failover to s2 instead of s1 because s1 is lagging I need evidence in the log to explain the behavior. >
Isn't it better to get this information via pg_stat_replication view (via state and sync_priority) columns?
We need the historical information to analyze and root cause in addition to the live debugging. It would be good to have better control over replication messages.