Master Master Replication ... do a fail over and a week agos datais revealed.

Hiya

I run MySQL Master - Master Replication. Ive had an interesting
situation whereby I failed over using heartbeat but whats is
interesting is that via the application (vbulletin), I see that the
forums was showing that a weeks ago data.

I had to re fail over to the original server and then the forums was up
to date again.

Has anyone seeing this type of issue with MySQL. I promise you that both
slaves are up and synced. I actually nagios monitor it.

If anyone could share some thought on MySQL's going on's or anything,
it would be appreciated.

Kind Regards
Brent Clark

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/mysql?unsub=gcdmg-mysql-2@m.gmane.org

Re: Master Master Replication ... do a fail over and a week agos datais revealed.

2010/10/18 Brent Clark :
> Hiya
>
> I run MySQL Master - Master Replication. Ive had an interesting situation
> whereby I failed over using heartbeat but whats is interesting =A0is that=
via
> the application (vbulletin), I see that the forums was showing that a wee=
ks
> ago data.
>

Why using M-M replication?
The book of "High performance Mysql" says it is not a reliable mechanism.

Regards.

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/mysql?unsub=3Dgcdmg-mysql-2@m.gmane.o rg

Re: Master Master Replication ... do a fail over and a week agos datais revealed.

--90e6ba6134e2c8db770492f4abd8
Content-Type: text/plain; charset=ISO-8859-1

On Tue, Oct 19, 2010 at 9:48 AM, short cutter wrote:

> 2010/10/18 Brent Clark :
> > Hiya
> >
> > I run MySQL Master - Master Replication. Ive had an interesting situation
> > whereby I failed over using heartbeat but whats is interesting is that
> via
> > the application (vbulletin), I see that the forums was showing that a
> weeks
> > ago data.
> >
>
> Why using M-M replication?
> The book of "High performance Mysql" says it is not a reliable mechanism.
>

There's various reasons why - almost all my setups also use it. It's not a
reliable mechanism for dual-active setups, but as a hot standby there's
nothing wrong with it whatsoever. Read the book again :-)

I don't have a straight explanation about why the secondary master offers
data from a week ago, though. If replication is running, maybe there's
something going on with the binlogging on the primary ? Check the primary's
master status and the secondary's slave status; check what's in the
primary's binlogs and in the secondary's relay logs; if need be check the
traffic that goes over the replication interface.


--
Bier met grenadyn
Is als mosterd by den wyn
Sy die't drinkt, is eene kwezel
Hy die't drinkt, is ras een ezel

--90e6ba6134e2c8db770492f4abd8--

Re: Master Master Replication ... do a fail over and a week agos data is revealed.

Johan,

You state that master - master is not reliable in dual active environments.
I am in the process of setting up just such an environment (moderate active
on the primary server, lighter activity on the other server.) Do you know
where I can get some information on the risks?

Thanks,

Carl

----- Original Message -----
From: "Johan De Meersman"
To: "short cutter"
Cc: "Brent Clark" ;
Sent: Tuesday, October 19, 2010 5:10 AM
Subject: Re: Master Master Replication ... do a fail over and a week agos
data is revealed.


> On Tue, Oct 19, 2010 at 9:48 AM, short cutter wrote:
>
>> 2010/10/18 Brent Clark :
>> > Hiya
>> >
>> > I run MySQL Master - Master Replication. Ive had an interesting
>> > situation
>> > whereby I failed over using heartbeat but whats is interesting is that
>> via
>> > the application (vbulletin), I see that the forums was showing that a
>> weeks
>> > ago data.
>> >
>>
>> Why using M-M replication?
>> The book of "High performance Mysql" says it is not a reliable mechanism.
>>
>
> There's various reasons why - almost all my setups also use it. It's not a
> reliable mechanism for dual-active setups, but as a hot standby there's
> nothing wrong with it whatsoever. Read the book again :-)
>
> I don't have a straight explanation about why the secondary master offers
> data from a week ago, though. If replication is running, maybe there's
> something going on with the binlogging on the primary ? Check the
> primary's
> master status and the secondary's slave status; check what's in the
> primary's binlogs and in the secondary's relay logs; if need be check the
> traffic that goes over the replication interface.
>
>
> --
> Bier met grenadyn
> Is als mosterd by den wyn
> Sy die't drinkt, is eene kwezel
> Hy die't drinkt, is ras een ezel
>


--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/mysql?unsub=gcdmg-mysql-2@m.gmane.org

Re: Master Master Replication ... do a fail over and a week agos datais revealed.

--0016369c897ab880d00492f722bd
Content-Type: text/plain; charset=ISO-8859-1

On Tue, Oct 19, 2010 at 1:03 PM, Carl wrote:

> Johan,
>
> You state that master - master is not reliable in dual active environments.
> I am in the process of setting up just such an environment (moderate active
> on the primary server, lighter activity on the other server.) Do you know
> where I can get some information on the risks?
>

On the MySQL site, I guess :-) Your major issue is likely to be the lack of
two-phase commits (which has been added in the most recent release, iirc):
even if you properly interleave your autoincrements, the lack of a true
global lock means you still risk conflicting updates.

I can't really tell you a single spot where I got my knowledge on the
subject - I'm a long-time and multi-platform DBA, so it's a combination of
stuff I read and an understanding of how things work - or don't.

--
Bier met grenadyn
Is als mosterd by den wyn
Sy die't drinkt, is eene kwezel
Hy die't drinkt, is ras een ezel

--0016369c897ab880d00492f722bd--

Re: Master Master Replication ... do a fail over and a week agos datais revealed.

To Clarify this a bit: You can only reliably do writes to one server.
Your reads can go to both servers with no problems.
For reads that need to read data that was just written though, you
need to read from the 'active' master, as you can not rely on
replication to be instantaneous.

On Tue, Oct 19, 2010 at 20:07, Johan De Meersman wrote=
:
> On Tue, Oct 19, 2010 at 1:03 PM, Carl wrote:
>
>> Johan,
>>
>> You state that master - master is not reliable in dual active environmen=
ts.
>> I am in the process of setting up just such an environment (moderate act=
ive
>> on the primary server, lighter activity on the other server.)  Do y=
ou know
>> where I can get some information on the risks?
>>
>
> On the MySQL site, I guess :-) Your major issue is likely to be the lack =
of
> two-phase commits (which has been added in the most recent release, iirc)=
:
> even if you properly interleave your autoincrements, the lack of a true
> global lock means you still risk conflicting updates.
>
> I can't really tell you a single spot where I got my knowledge on the
> subject - I'm a long-time and multi-platform DBA, so it's a combination o=
f
> stuff I read and an understanding of how things work - or don't.
>
> --
> Bier met grenadyn
> Is als mosterd by den wyn
> Sy die't drinkt, is eene kwezel
> Hy die't drinkt, is ras een ezel
>



--=20
Walter Heck
Founder @ OlinData (http://olindata.com)
Co-founder @ Tribily (http://tribily.com)

--
MySQL General Mailing List
For list archives: http://lists.mysql.com/mysql
To unsubscribe: http://lists.mysql.com/mysql?unsub=3Dgcdmg-mysql-2@m.gmane.o rg

Re: Master Master Replication ... do a fail over and a week agos datais revealed.

--90e6ba3fd209db1dc004931a6c55
Content-Type: text/plain; charset=ISO-8859-1

On Wed, Oct 20, 2010 at 1:37 AM, Walter Heck - OlinData.com <
lists@olindata.com> wrote:

> To Clarify this a bit: You can only reliably do writes to one server.
>

Also not *entirely* true: nothing prevents you from using the two masters
for distinct databases - or even tables - and just having them as mutual
hot-standby.

It's very important to realise that replication does NOT scale your write
capabilities, though: every write that happens on one side must also happen
on the other side.


--
Bier met grenadyn
Is als mosterd by den wyn
Sy die't drinkt, is eene kwezel
Hy die't drinkt, is ras een ezel

--90e6ba3fd209db1dc004931a6c55--