>From: Claudio Nanni [mailto:claudio.nanni@stripped]
>Sent: Wednesday, April 21, 2010 2:12 AM
>Subject: Re: better way to backup 50 Gig db?
[JS] Unless I've forgotten something from earlier in my career (what day is
it, anyways?), there are three aspects to this problem:
1. Ensuring that your databases, slave and master individually, are internally
2. Ensuring that your master has captured the latest externally-supplied data;
3. Ensuring that your slave and you master are totally in synch.
#1 is the proper goal for the master. That's the whole point of ACID. For the
master database, #2 is unattainable. You can buffer as many times and as many
ways and as many places as you like, there is always going to be the
**possibility** that some incoming data will be lost. Even if you push the
problem all the way back to a human user, it will still be possible to lose
data. If something is possible, it will happen: perhaps not for millennia, but
more likely as soon as you leave on vacation.
Similarly, #1 is an attainable and necessary goal for a slave; and #2 is just
as unattainable for a slave as for a master. The only way to guarantee #3 is
to include the replication somewhere in the ACID transaction. The penalty for
that is going to be a loss of throughput, possibly a horrendous loss of
throughput. That is where somebody needs to do a cost/benefit analysis.
>Just my two cents
[JS] ... and mine ...
>Gavin Towey wrote:
>You can make binary backups from the master using filesystem snapshots. You
>only need to hold a global read lock for a split second.
>MySQL General Mailing List
>For list archives: http://lists.mysql.com/mysql
>To unsubscribe: http://lists.mysql.com/mysql?unsub=1