Thu Jan 13 06:19:26 PST 2005
- Previous message: [Slony1-general] current slony1 limitations
- Next message: [Slony1-general] current slony1 limitations
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
...and on Wed, Jan 12, 2005 at 10:33:14AM -0800, Dan Hrabarchuk used the keyboard: > I've never seen "corrupted" replication on MySQL, but I've only ever > used it on rather lite sites. > > The replication is much lighter on the system though, as it's > transaction log based. > > Also setting up a new slave is MUCH faster, but you need to make the > initial copy with a downed server. > > Dan > Hello, Dan. I'm afraid that's not true. While being the fastest and most "elegant" solution, there is most certainly the option of saying LOAD DATA FROM MASTER or LOAD TABLE tbl_name FROM MASTER which will populate the new slave with data from the tables directly, rather than using transaction logs. It works only with MyISAM tables though, and it acquires a global read lock on the master, which means no updates are possible during the time of the load, so it is essentially rather useless, unless you temporarily set master to a slave replica and later on reset slave counters and update master to the real one. My experience with MySQL replication (we had a number of 100+GB databases where I used to work) is that it is consistant with "the MySQL way" of doing things - simple for the straightforward tasks, next to impossible for complex stuff. It is, if you leave it running and not touch it, rather reliable, for as long as you optimize both the master and the slaves to be reasonably in sync with reality, but it may soon become a pain, as MySQL is not too fool-proof, and nothing prevents you from, for example, deleting a master log while it is still in use by slaves. I'm afraid I haven't had sufficient experience with Slony-I so far, so I'm not able to compare the two. From what I've seen this far though, my gut feeling is that Slony-I is a different beast alltogether, regarding both its initial goal (as providing a number of "nodes" of which any one can be a master for a set data source at any given moment in time) as well as its overall flexibility. HTH, -- Grega Bremec gregab at p0f dot net -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available Url : http://gborg.postgresql.org/pipermail/slony1-general/attachments/20050113/259a4012/attachment.bin
- Previous message: [Slony1-general] current slony1 limitations
- Next message: [Slony1-general] current slony1 limitations
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the Slony1-general mailing list