Re: Replication basics

[ian douglas]

1. If a new database is created on db1, this is not replicated on db2
and db3. ... is there a better way to automate duplicating this 
>>database on the slaves?
Yes put in your my.cnf on the master binlog-ignore-db=mysql,test this
means the master will replicate any event to any database other then mysql,
test.
So, by your statement:
        binlog-ignore-db=
... and leaving it blank, will tell MySQL to replicate EVERY database? 
That's the behavior that I want. I will not be ignoring any databases on 
the master - all data on the master (databases, tables, data) must exist 
on every system, at least for the time being. Later on, we may replicate 
only certain tables onto certain slaves depending on our user load and 
which databases are more frequently used. However, for the foreseeable 
future, all data must exist on all machines.

2. If a database exists on all three machines already and I create a
table on db1, this is also not replicated on db2 or db3.
Why not?
That's my question too. I haven't seen anything on how to allow this 
behavior yet. I've seen some text about "LOAD TABLE tablename FROM 
MASTER" but I need a way to automate this so there is less manual 
intervention required.

Are you explicitly replicating a set of tables from your slaves via
replicate-do-table=foo.bar?
I need to replicate everything, and I'm not setting any options like 
this in the slave's my.cnf. Each slave must be a complete mirror copy of 
the master.

3. I haven't checked yet whether "ALTER TABLE ..." commands have worked
across the replication.
Yes they do.
To finish my statement: "... in my implementation" - that is, I haven't 
tested my implementation to know if an "ALTER TABLE ..." query works or not.

4. Darn RedHat and their logrotate utility: ... created db1.001, 
>>db1.002, db1.003, etc., ... This also happens if db1 reboots.
Stop the logrotate script.
Which I did last week when I realized this was why my slaves hadn't 
sync'd up in a week... But this past weekend a few transformers blew in 
the neighborhood around our office and a power surge shut off db1 
(master) - when it came back online, it created a db1.002 file and 
updated its own db1.index file, yet the two slave machines did not 
update their master.info records. How can I automate the slaves to 
detect this change, and use the db1.002 file? (db2 and db3 were not shut 
off, they kept running just fine)

5. In the event of power failure, or system failure, if db1 is offline,
how can I set the systems so either db2 or db3 becomes a master?
write your own load / failover software.
Thanks for the pointer. Was hoping there was some sort of round-robin 
scenario.

6. [paraphrase] Rewriting Perl and PHP code
Look at DBI::Multiplex
... anyone know of a PHP implementation of this? It would sure save me 
some time.

7. Finally, would 'fake replication' work if the MySQL database files
were on a RAID system and mounted via NFS 
No, not reliable.
Noted, thank you for saving me some grief.
-id
--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe:    http://lists.mysql.com/[EMAIL PROTECTED]