Recommended replication solution?

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Hello everyone,

Yes I know, "size one not fits all", I specifically need:

* preferably multi-master

* local read, cluster-wide write

* a solution that allows me to maintain consistency between masters in
case of single node failure

* last but not least, production-ready

Load balancing is not important in my case, neither is connection
pooling. Fault-tolerance would be nice to have, but it's not absolute
must - what concerns me most is cluster-wide data integrity, and not
necessarily HA.

I need to put the replicated database behind SQLAlchemy, and the
simplest setup would be if SA were handing queries over to a cluster as
if it were a single local db.

I'd love to use Postgres-R, as its explicit design goals seem most
sensible, but they say on their page that it's not production-ready yet.
(having said that, has anybody used that in production of some kind?)

Londiste seems complicated to set up and I was not able to gather
specific information on whether it is able to meet the requirements above.

Pgpool-II seems viable, and slide 17 on
http://www.pgcon.org/2008/schedule/attachments/66_pgpool-II-demo.pdf
seems to suggest it may be used as multi-master, but I wasn't able to
figure out setup for this yet.

Anybody has production experience with those? Pros? Cons?

- --

Regards,
mk

- --
Premature optimization is the root of all fun.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/

iQEcBAEBAgAGBQJNC3ZtAAoJEFMgHzhQQ7hOMgQH/0PTxhlGjuqg9Z0cKjsRutWI
02ipZ5ve2OhoCVAtgdje5w5Ts4bNf5l4dn0AmQ6uWSpruBwpuRNIp2tvpAIpc9q/
+NiLtgH5uSdrGkKIgJd53sGdtLqUz1Ax+n0MHD+9IRjmN65pbuPaxzgPAQ4z3Lxx
9HvKanzBT9VxT0IXAM7OIWrhuCsDh4DT89/JrRiBlPwWq823rixI6QBc3QWAda+I
P3NYs9Dcg5nRSZQF8VV66otGxWj1aDOu0maOHuIUX0A2C6MAM3dvzIovxV3SBEeh
zAV2t7TnTrGyCAftZCOxe1c6jQ3pAGPbaOHbyscQ/d3Rm/FpcR/LxY4RhwH5/Fs=
=7jMW
-----END PGP SIGNATURE-----