[Pgcluster-general] FYI: PGCluster-1.9.0rc1 released

[Rick Vernam]

On Wednesday 06 February 2008 03:16:40 am Tommi Berg wrote:
> Hi,
>
> At.Mitani wrote:
> > It would be a grate pleasure if you could send the bug reports.
>
> Exept VACUUM ANALYZE I have already posted these as bugs for 1.7.0rc8,
> but I tested that these all exists also at 1.9.0rc1.
>
> - Running VACUUM ANALYZE on one Cluster DB blocks all writes to that
> table. This leads rather nasty things with autovacuum and large tables.
> - Using placeholders/bind params ("dollar sign numbers") and running
> writes at two or more threads will cause clusters to hang.
> - Load balancer only connects to two first clusters.
> - Recovery fails with "large" database, timeout after message: "1st
> sync_table_space OK".
I encounted this as well.  After finding a good bwlimit for rsync, I no longer 
experience this.

> - Using inet_server_addr() (might be more functions like this?) at
> writes leads to wrong data.
>
> Last, I don't know if this is a bug or feature, the write performance
> with three cluster db servers with dedicated replication server is about
> three times slower than using single postgres server.
>
>
> Tommi Berg
> _______________________________________________
> Pgcluster-general mailing list
> Pgcluster-general at pgfoundry.org
> http://pgfoundry.org/mailman/listinfo/pgcluster-general