Clustering mysql: MySQL-CLuster or Replication? Your opinion
Yukari
up a cluster for my mysql-servers.
And for those who didn't: I got the advice to try either MySQL-Cluster
(see http://www.mysql.com/products/database/cluster/) or replication
(see http://dev.mysql.com/doc/refman/5.0/en/replication.html).
Well, I'm a newbie to all of them. In addition to that I'm a newbie at
linux, too.
That's why I was a little bit afraid of using linux at the beginning
and wanted to stick with windows. But then MySQL-CLuster would drop
out, because it's not recommended for use on windows.
I have two (or actually three) options now and it's very difficult for
me to decide which one I should choose, because of the reasons
mentioned before. That's why I'm asking you. Which one would YOU
prefer and WHY?
Using MySQL-Cluster or replication (regardless which os used for the
second... linux or windows)?
I keep on reading stuff for both of them to find my best way myself
but maybe you can help me with your experiences.
Yukari
> Maybe some of you already read in my other thread that I want to set
> up a cluster for my mysql-servers.
>
> And for those who didn't: I got the advice to try either MySQL-Cluster
> (seehttp://www.mysql.com/products/database/cluster/) or replication
>
> Well, I'm a newbie to all of them. In addition to that I'm a newbie at
> linux, too.
> That's why I was a little bit afraid of using linux at the beginning
> and wanted to stick with windows. But then MySQL-CLuster would drop
> out, because it's not recommended for use on windows.
>
> I have two (or actually three) options now and it's very difficult for
> me to decide which one I should choose, because of the reasons
> mentioned before. That's why I'm asking you. Which one would YOU
> prefer and WHY?
>
> Using MySQL-Cluster or replication (regardless which os used for the
> second... linux or windows)?
>
> I keep on reading stuff for both of them to find my best way myself
> but maybe you can help me with your experiences.
I forgot to mention: I'm searching for the BEST solution and also for
the EASIEST solution at the same time. Difficult combination, I know.
But maybe some compromise is possible and I just don't see it yet.
That's why I'm asking you.
Axel Schwenke
>
> I have two (or actually three) options now and it's very difficult for
> me to decide which one I should choose
You did not give enough data to assist in a decision. But based on
your questions, probably none of the two mentioned technologies
will solve your problem. Even less be the best solution.
My answer in <6m3dt6-...@xl.homelinux.org> is pretty exhaustive.
So why don't you go and read about cluster manager software?
EOD for me.
XL
Yukari
> Yukari <yukaricha...@googlemail.com> wrote:
>
> > I have two (or actually three) options now and it's very difficult for
> > me to decide which one I should choose
>
> You did not give enough data to assist in a decision. But based on
> your questions, probably none of the two mentioned technologies
> will solve your problem. Even less be the best solution.
>
> So why don't you go and read about cluster manager software?
>
> EOD for me.
>
> XL
Hm, ok so what additional data would you need?
Your answer in... where?
Hehe, yeah well I'm already reading much stuff about the different
options, cluster manager you posted before is also on my list.
Maybe I forgot to put it on the list here, because I'm not there yet
and don't have a clear enough imagination of it.
So, at this point you would say a Cluster Manager would be the most
suitable option? But which one and why? There are also many.
The Natural Philosopher
I think his point is, what are you finally trying to achieve?
Resilience?
High data throughput?
A system that will survive being stolen?
(No joke: one arrest made when person using datacenter computer suddenly
went off line. Phoned sysdamin, who looked in..and was just in time to
get the number plate of the car speeding away with aforesaid computers)
Often you can solve things another way. For example one very high spec
machines is actually better/faster/cheaper than three lower power ones. Etc.
WE went through this sort of stuff to try and build rock solid mail
servers for 10,000 customers. In the end, hot swap RAID disk subsystem,
massive RAM and a big processor, and enough spares to fix it on site,
and a good engineer on call, worked out best. Plus some useful tools to
monitor and alert if memory, processes or disk space suddenly escalated. .
Yukari
wrote:
The idea behind everything was that the system (the databases) is
available anytime. Actually not really my idea, I'm just the poor
devil who has to find out how to realize it ;)
No, back to the topic: If anything fails, technical, human or whatever
issues so that one machine can't go on working properly the other one
should resume it's business.
If there's only one machine and anything happens to THAT, the whole
service would be down and I wanted to prevent that.
I don't know if it even would be possible to use one big machine,
because as far as I know most of the hardware is already there and has
to be reused.
Another precondition that would also be nice to achieve would be easy
monitoring later.
It's also planned to have a NLB-system in front of the mysql-cluster,
where the webapplications will be installed on, but I don't know if
this matters here.