On Tue, Oct 18, 2011 at 6:39 PM, Samuel Hwang <samuel@replicon.com> wrote:
> The log is getting from PostgreSQL 9.0.4
> Basically we set up streaming replication hot-standby slave while master is
> under heavy load
> The slave started but not accepting read-only queries,
> every request will trigger the "FATAL: the database system is starting up"
> error.
>
> The slave will eventually be able to accept read-only queries after the load
> on master is removed.
> We don't see the same problem if we set up hot-standby slave while master is
> not under heavy load.
>
> the logs follow:
> 2011-10-18 10:34:06 MDT [17570]: [13-1] LOG: consistent state delayed
> because recovery snapshot incomplete
Caveat #2 applies here
http://developer.postgresql.org/pgdocs/postgres/hot-standby.html#HOT-STANDBY-CAVEATS
The consistent state is delayed until your long running transactions
end, which is workload dependent but transient.
It's possible we will find another way of doing this in a future release.
Until then, I suggest starting base backup to create the standby when
not running both long transactions and transactions with many
subtransactions.
--
Simon Riggs http://www.2ndQuadrant.com/
PostgreSQL Development, 24x7 Support, Training & Services