Bonjour Sasha,
* Will it work only with 2 servers or can there be many slave of a
master server ?
* How will we ask for a table to be a replication of a distant one ?
In the create statement ? Is it TABLE or DATABASE replication ?
* Will this replication process sync the tables after the master
issues a LOAD DATA INFILE ?
* You are talking about 'updates' on the slave. If the masters loads
a lot of data in it's table (LOAD DATA IN FILE form example), will
the slave have to do updates rows by rows or will it be able to
replicate as quickly as the master did it's LOAD DATA INFILE ?
Thanks in advise for enlightenment of these points.
Regards,
Alex.
smc> a) master server keeps update binary log
smc> b) slaves connect to it and issue COM_BINLOG_DUMP command, specifying the log
smc> file that they have read last, and the offset in it where they left off
smc> c) master keeps an index of logs, and gives a full dump of all the logs to the
smc> slave up to the current one. When eof is reached, the master blocks and starts
smc> feeding updates as they arrive to the slave
smc> d) the slave reads the updates, and executes them, keeping track of where it is.
smc> In case of problems the slave will always know where it left off, as it flushes
smc> the master info on every complete update
smc> e) special handling of insert_id, user_variables and timestamps to ensure
smc> consistency
smc> f) special command to the master server to dump the entire table in case the
smc> slave loses everything, or is has just been set up and starting from scratch
smc> I've already finished a), b) c), and am currently working on d) - I am quite
smc> confident a) through e) will make it into 3.23.15, and maybe even f)
