Hi.
I have Pg 9.0 wal shipping hot standby secondary server.
The primary is under constant stream of updates (avg 20 TXID/s).
There are many lengthy COPY FROM operations in the primary.
After every restart of secondary postgres I observe that it takes a
fair amount of time (sometimes few minutes, sometimes much more) to
replay new WAL logs achieve consistent state and start serving R/O
queries.
Even when R/O queries were served directly before the restart.
Why does postgres take so longto reach consistent state?
Can I ever try to impose a time limit on this? How?
What prevents postgres from using last consistent restart point from
before the cluster restart? This way I would not have to wait so long
after restart to serve R/O traffic.
Is this issue any different in 9.2?
Is this issue mitigated in any way if I switch on streaming replication?
thanks for any answers & suggestions.
Filip