Hi,
On 16-nov-2006, at 14:41, Leandro Borges wrote:
I think that the right setup is the second (I don't know if you
can setup the first configuration because each backend can be
access only by one controller).
Great, I was hoping that the second setup would be possible/preferrable.
I already read that backends cannot be shared by multiple controllers.
However, technically speaking that is not the case here. Only one of the
controllers is active at any time.
I currently use the second configuration, just like the diagram,
Linux Virtual Server -> Apache mod_jk -> Tomcat -> Sequoia, and if
one controller fails the driver notice and all queries go to the
remain controller.
Are you also using a RAIDb-1 setup? So I'm right that the RAIDb-1
replication
of (write-)queries is done by the controllers amongst themselves?
Thanks for your response!
Leander
Leander Koornneef escreveu:
Hi,
I'm currently designing a new setup where I'm thinking about
deploying
Sequoia. I've read some docs and skimmed the mailing list archive,
but
still have some questions about how Sequoia works and how to set
things
up properly. We're dealing with a clustered Java/Tomcat webapp,
running
on at least two physical servers. The webapp connects to a Postgresql
database, which must be replicated realtime and synchronous. The main
concerns here are availability and redundancy. Performance also
counts
of course, but is not nearly as important as availability and
redundancy.
So, we'll be setting up a RAIDb-1, with probably two physical
database
servers. My question is about how to setup the Sequoia controllers in
this case. I definitely don't want the controller to be a single
point of
failure. I made two diagrams of potential setups, which I'll
attach to this
email. However, I am not sure if these setups are even possible at
all.
I would appreciate it if someone could please comment on them.
The first setup (sequoia1.png) uses one active Sequoia controller
at any
time, which controls the RAIDb-1 over the two DB servers. The other
controller is stand-by. If the DB01server, or the DB01 Sequoia
controller
fails, something like Heartbeat will switch DB01's IP address to
the DB02
server and start up the Sequoia controller on DB02.
The second setup (sequoia2.png) uses two active Sequoia controllers
(under normal circumstances), which both control a single database
instance. The Sequoia/JDBC connection string should now list two
Sequoia controller backends, I suppose? The RAIDb-1 is now achieved
by communication between the two controllers, so every (write-)
query will
be replicated by sending it from one controller to the other. Is
this even
possible? And what will happen if one of the controllers/databases
fails?
Will the Sequoia JDBC driver notice this and redirect all queries
to the
remaining controller?
These questions are quite fundamental and probably also FAQ, but I
could not find a definitive answer in the docs and/or archives.
Basically, what I want to know is: are these two setups even
possible/feasible
and if so: which of them is "the best" in terms of availability
and redundancy?
thanks,
Leander
_______________________________________________
Sequoia mailing list
[email protected]
https://forge.continuent.org/mailman/listinfo/sequoia
