FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > CentOS > CentOS

 
 
LinkBack Thread Tools
 
Old 03-17-2010, 03:19 AM
"Neil Aggarwal"
 
Default MySQL max clustering package?

> Mysql by itself has built in "clustering" though
> there can be significant limitations in it depending on your
> requirements.

I agree. The built in cluster has too many limitations to
be useful, but MySQL master-master replication gives a very
good alternative to a true cluster. We use it to deploy
geographically redundant systems and it has worked very
well for us.

Neil

--
Neil Aggarwal, (281)846-8957, http://UnmeteredVPS.net/mysql
MySQL pre-installed on a virtual private server for $25/mo
Unmetered bandwidth = no overage charges, 7 day free trial

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 04:24 AM
John R Pierce
 
Default MySQL max clustering package?

Neil Aggarwal wrote:
> I agree. The built in cluster has too many limitations to
> be useful, but MySQL master-master replication gives a very
> good alternative to a true cluster. We use it to deploy
> geographically redundant systems and it has worked very
> well for us.
>


master-master replication has all kinds of inherent issues if you're
concerned with data and transactional integrity. its pretty hard to
uncommitted a transaction some time after its been committed.

if its done fully safely, its incredibly slow, as you need global
locks. otherwise, you're playing russian roulette and relying on your
data access patterns to inherently avoid conflicts. if your problem
space can be partitioned geographically, then its doable.



_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 05:42 AM
"Neil Aggarwal"
 
Default MySQL max clustering package?

John:

> master-master replication has all kinds of inherent issues if you're
> concerned with data and transactional integrity.

We have evaluated the problems and think we have found adequate
workarounds for them. The sites we deploy are e-commerce sites so:
1. There are a lot more reads than writes.
2. When we need to write to the database, it is saving
info for an order. All orders are independent
of each other so there is no conflict for those.
3. Product updates, fulfillments, etc. happen once
daily at night when the traffic is almost nonexistant.
The updates are sent to only one server so there is no
conflict there.

The only potential place a conflict may occur is in
the qty available for a specific product. The inventory
system updates the inventory regularly so even if the number
is wrong, it gets refreshed shortly thereafter.

We even built an application layer on top of master-master
replication to handle cases where a transaction fails.

We are using this system for several large clients
and it is working well. Being able to have geographical
redundancy at a reasonable cost (A true cluster would
be very high cost) outweighs the limitations.

Thanks,
Neil

--
Neil Aggarwal, (281)846-8957
MySQL pre-installed on a virtual private server for $25/mo
Unmetered bandwidth = no overage charges, 7 day free trial

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 10:29 AM
Alexander Georgiev
 
Default MySQL max clustering package?

2010/3/17 Neil Aggarwal <neil@jammconsulting.com>:

> The only potential place a conflict may occur is in
> the qty available for a specific product. *The inventory
> system updates the inventory regularly so even if the number
> is wrong, it gets refreshed shortly thereafter.
>

Do you mean that a separate job, iterates the orders, accumulates the
real ordered quantity and subtracts it from some "initial quantity" in
order to produce available quantity?

What do you do in cases where you have oversold a product. I mean when
the "ordered quantity" got bigger than the "available quantity" due
to a conflict in available quantity field? I assume that the system
sends an email to the warehouse to increase additionally the quantity
of that product?

> We even built an application layer on top of master-master
> replication to handle cases where a transaction fails.
>

Could you describe a case where a transaction has failed , and how you
deal with it?

Thank you!

Alex
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 11:16 AM
JohnS
 
Default MySQL max clustering package?

On Tue, 2010-03-16 at 23:19 -0500, Neil Aggarwal wrote:
> > Mysql by itself has built in "clustering" though
> > there can be significant limitations in it depending on your
> > requirements.
>
> I agree. The built in cluster has too many limitations to
> be useful, but MySQL master-master replication gives a very
> good alternative to a true cluster. We use it to deploy
> geographically redundant systems and it has worked very
> well for us.
>
> Neil

Well what are your plans when it gets the AXE??

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 11:27 AM
JohnS
 
Default MySQL max clustering package?

On Wed, 2010-03-17 at 13:29 +0200, Alexander Georgiev wrote:
> 2010/3/17 Neil Aggarwal <neil@jammconsulting.com>:
>
> > The only potential place a conflict may occur is in
> > the qty available for a specific product. The inventory
> > system updates the inventory regularly so even if the number
> > is wrong, it gets refreshed shortly thereafter.
> >
>
> Do you mean that a separate job, iterates the orders, accumulates the
> real ordered quantity and subtracts it from some "initial quantity" in
> order to produce available quantity?
>
> What do you do in cases where you have oversold a product. I mean when
> the "ordered quantity" got bigger than the "available quantity" due
> to a conflict in available quantity field? I assume that the system
> sends an email to the warehouse to increase additionally the quantity
> of that product?
>
> > We even built an application layer on top of master-master
> > replication to handle cases where a transaction fails.
> >
>
> Could you describe a case where a transaction has failed , and how you
> deal with it?

Yes im interested to hear that also because im not aware that MySQL has
any type of feature like:
database.rollback() or table and collumn rollback on a bad transaction.
I have always heard the replication of MySQL could not keep up with lots
of writes.

Have you thought of separating the databases? One for the reads and one
for the write on different raids? Despite what some may believe this
can be done.

John

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 12:10 PM
Chan Chung Hang Christopher
 
Default MySQL max clustering package?

JohnS wrote:
> On Tue, 2010-03-16 at 23:19 -0500, Neil Aggarwal wrote:
>>> Mysql by itself has built in "clustering" though
>>> there can be significant limitations in it depending on your
>>> requirements.
>> I agree. The built in cluster has too many limitations to
>> be useful, but MySQL master-master replication gives a very
>> good alternative to a true cluster. We use it to deploy
>> geographically redundant systems and it has worked very
>> well for us.
>>
>> Neil
>
> Well what are your plans when it gets the AXE??
>

firebirdsql of course.
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 12:41 PM
JohnS
 
Default MySQL max clustering package?

On Wed, 2010-03-17 at 21:10 +0800, Chan Chung Hang Christopher wrote:
> JohnS wrote:
> > On Tue, 2010-03-16 at 23:19 -0500, Neil Aggarwal wrote:
> >>> Mysql by itself has built in "clustering" though
> >>> there can be significant limitations in it depending on your
> >>> requirements.
> >> I agree. The built in cluster has too many limitations to
> >> be useful, but MySQL master-master replication gives a very
> >> good alternative to a true cluster. We use it to deploy
> >> geographically redundant systems and it has worked very
> >> well for us.
> >>
> >> Neil
> >
> > Well what are your plans when it gets the AXE??
> >
>
> firebirdsql of course.
You have a loving for Pontiacs?

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 01:01 PM
"nate"
 
Default MySQL max clustering package?

JohnS wrote:

> I have always heard the replication of MySQL could not keep up with lots
> of writes.

I don't think MySQL replication has an issue with number of writes,
at least with regular replication(can't speak to multi master stuff),
all replication is is the DB sending the raw query to the other
system to execute, so provided the other system(s) your replicating
to are at least as fast as the system doing the real work the others
should have no trouble keeping up.

nate


_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 
Old 03-17-2010, 02:56 PM
"Neil Aggarwal"
 
Default MySQL max clustering package?

> Well what are your plans when it gets the AXE??

We will probably consider Maria DB. Hopefully,
it will be mature enough by then.

Neil

--
Neil Aggarwal, (281)846-8957, http://UnmeteredVPS.net/centos
Virtual private server with CentOS 5.4 preinstalled for $25/month!
Unmetered bandwidth = no overage charges, 7 day free trial

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos
 

Thread Tools




All times are GMT. The time now is 04:28 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org