Yesterday was Northwest Python Day. All of the talks were great, and the day was immensely rewarding and enjoyable.
Andrew introduced me to Andy McKay, with whom I’ve traded a bit of e-mail but never met. It’s an oddity of today’s world that you can do business, or strike up a friendship via a few e-mail exchanges, with someone, without ever having actually met them.
Most of the attendees brought some food, and most of them brought breakfast food, and most of the breakfast food were doughnuts. We were swimming in doughnuts! Boy oh boy, there were a lot of doughnuts.
I left right before the last talk, as I was beginning to nod off. We were in one room for the entire day, sitting in the same chairs… I started fading by by 4:15 pm, and concluded that the better part of valor would be to leave at that point.
I particularly liked Joshua Drake‘s talk on Replicator for PostgreSQL. I’ve been thinking about database replication for a large Plone project at work (which is now on hold, but that’s a story for another day), primarily for redundancy, although the additional read performance would be nice. We’d be adequately served by two master-slave dbs. Open-source alternatives for this include ZEO RAID, and something like DirectoryStorage, with the underlying files replicated via a replication tool.
But given the importance of this particular database, I’ve provisionally selected Zope Replication Services (ZRS), a commercial product. It would provide a single read-write primary server and multiple read-only secondary servers, with automatic failover and promotion. As much as I support open-source, if database replication isn’t done right, you will have Hell to pay, and go insane and blind.
But I want to stay aware of alternatives here, whether for OODB or RDB tiers. And the $16K ZRS licensing fees would be nice to eliminate…if I could be sure that an open-source alternative would not send me to Hell. At any rate, Replicator sounds like a good tool for PostgreSQL replication, and as I hadn’t heard of it before, the talk was a pleasant surprise.
John,
Before you dump cash on ZRS, you may want to look at RelStorage, which lets you store your Zope pickles in Oracle, Postgres or MySQL, and take advantage of those systems’ existing replication architectures.
Thank you for the tip!
Also look at ZEORaid, which does the same thing as ZRS but is open source.
Note that currently the only working replication setup for relstorage is Oracle.