asterisk users - Apr 2007 - Re: Voicemail on Different Server (MySQL Replication split thread)

> Your experience with database replication is not unique. I have seen
> this happen with many flavours of database, not just MySQL. At the
> "critical" sites where I've worked, database replication is
not even on
> the table as an option for precisely the reasons you state above: I have
> yet to meet someone else who has had a positive experience with it.
I'm using MySQL replication for my realtime database, 1 master with
10+ slaves.  I have not had any data corruption or problems since
implementation 6 mos ago.  To qualify that I must say, I only do
one-way replication. Master and Slaves are on the same switch
fabric,
>.7msec latency between hosts.  I attended a MySQL HA class,
Replication was covered in depth and discussed heavily, at no time did
I receive cautionary information about possible data corruption or bad
experiences from others in the class or from the instructors.

I do daily backups of the master database and also regularly check bin
file status between all servers to ensure no server is falling behind.

I'm not trying to dispute or start a flame war, I'm sure replication
is not perfect and 100% reliable for every instance.  I'm sure there
are many stories of failed replication or data corruption when
replication is not implemented properly or setup in an environment not
particularly suited well for replication.  Just wanted to add my own
experience.

JR
-- 
JR Richardson
Engineering for the Masses

JR Richardson wrote:
>> Your experience with database replication is not unique. I have seen
>> this happen with many flavours of database, not just MySQL. At the
>> "critical" sites where I've worked, database replication
is not even on
>> the table as an option for precisely the reasons you state above: I
have
>> yet to meet someone else who has had a positive experience with it.
> 
> I'm using MySQL replication for my realtime database, 1 master with
> 10+ slaves.  I have not had any data corruption or problems since
> implementation 6 mos ago.  To qualify that I must say, I only do
> one-way replication. Master and Slaves are on the same switch fabric,
>> .7msec latency between hosts.  I attended a MySQL HA class,
> Replication was covered in depth and discussed heavily, at no time did
> I receive cautionary information about possible data corruption or bad
> experiences from others in the class or from the instructors.
Having master and slave servers in the same switch fabric is the only
situation in which I would consider replication.

The cases that I described were with machines in separate subnets.
Replication simply doesn't work that well when there is significant
latency. Did they mention that in your HA class?
> I do daily backups of the master database and also regularly check bin
> file status between all servers to ensure no server is falling behind.
> 
> I'm not trying to dispute or start a flame war, I'm sure
replication
> is not perfect and 100% reliable for every instance.  I'm sure there
> are many stories of failed replication or data corruption when
> replication is not implemented properly or setup in an environment not
> particularly suited well for replication.  Just wanted to add my own
> experience.
Thanks,

-Stephen-