>> > The scenario I'm most interested in is this:
>> >
>> > 2 servers - a master and a hot standby. All writes are
>> sent to master, reads are split between master and hot
>> standby.
>> >
>> > 1) If the hot standby goes down, how do I redirect
>> reads to the master?
>>
>> pgpool-II 3.0 will take care of this.
>>
>> > 2) If the master fails
>> >,A (B ,A (B-how do I automatically
>> promote the standby to master and send all reads/writes to
>> the new master?
>>
>> This is covered by pgpool-II 3.0 as well.
>>
>> >,A (B ,A (B-what happens when the old
>> master comes back up? Do I need to so anything to make it
>> catches up to the new master?
>>
>> I recommend to use it a standby. Such a configuration is
>> possible by
>> using pgpool-II 3.0.
>> --
>
> Oh so I'd still need a proxy such as pgpool-II for HA setup?
> I was thinking that with the new built-in replication in 9.0 there would be no need to use pgpool-II.
PostgreSQL 9.0's replication still lacks automated failover/load
balance/query dispatching(send read/write query to primary, send read
query to standby). So if you need these, you would want to use
pgpool-II or any other proxy solutions.
> If pgpool is still necessary why not also use it for replication? What would be the advantages of using the 9.0's
built-inreplication as opposed to pgpool's replication?
Each replication solution has its own merit/demerit. For example, if
you need synchronous replication, pgpool-II is for you. If you are ok
with async, PostgreSQL's replication is quite nice.
So it depends on you.
--
Tatsuo Ishii
SRA OSS, Inc. Japan
English: http://www.sraoss.co.jp/index_en.php
Japanese: http://www.sraoss.co.jp