Does anyone know if zabbix works with pgpool-II with replication, load balancing and maybe even parallel queries? My db is under heavy fire from lots of users viewing lots of graphs&screens and distributing the SELECTs would help a lot, but the restrictions list is pretty intimidating.
-
-
Thanks, but I'd rather not mess around with the db structure.
I have to serve lots of screens that have many graphs that have many items, over various periods of time, all of this translating to a whole bunch of almost random SELECTs sent at the server at the same time.
DB is about 200GB now (using pg8.4.8), and the 4 discs in RAID10 just can't cope anymore. I could go with faster storage devices, but it'd be cheaper to add a secondary backend and keep everything together with pgpool-II. This would also give me an easy way to expand in the future if/when needed, instead of starting from scratch with sql analysis and it comes with the implicit added benefit of having online redundancy for when one of the dedicated db servers dies.Comment
-
I think pgpool-II probably would help you to scale out read only transactions. Keep in mind though, if in a transaction you do write, and then do a read, it won't balance the read, it will have to go to the master.
If writes are not a problem (you don't have that many) and only reads are, you could use pgpool-II's synchronous replication mode. Otherwise, if you could live with a little latency, consider using pgpool-II with PostgreSQL 9.0 and its new replication features.
You can also keep your eye on another maturing project, Postgres-XC. It is a scale out solution with cluster-wide consistency. (Disclaimer: I am involved in the project.) If you do experiment with it, I would recommend first trying with the latest HEAD, not the last release.
Regards,
MasonComment
Comment