The more I know about having a farm of servers, the more worried I get
about them:
Farm Cons:
* No way to sync the global scopes (Application, Server)
* No CFCs in session scope (Has this changed in CFMX7? Guess I can
look it up.)
* Only native way to do session sharing is to
There are plenty of pros that come along with a cluster. Namely speed,
high-availability and ease of patch management.
-Adam
On 6/23/05, jacksonj @ calib. com jacksonj @ calib. com
[EMAIL PROTECTED] wrote:
The more I know about having a farm of servers, the more worried I get
about them:
Is it possible to share sessions without clustering?
(I'd like to be able to know that when I'm hitting Server A, that the
application I'm seeing is coming from Server A. This is for hardware
load-balancing and fail-over reasons.)
Here's a real world example: Sometimes, I need to reload
by CFMX cluster you are speaking of SOFTWARE based clustering in JRun, not a
load balancer correct? Try this code below, it will tell you what instance
you are on. I use this. Its a real PITA though to make sure you are on
instance foo or goo.
!DOCTYPE HTML PUBLIC -//W3C//DTD HTML 4.01
by CFMX cluster you are speaking of SOFTWARE based clustering in JRun, not a
load balancer correct? Try this code below, it will tell you what instance
you are on. I use this. Its a real PITA though to make sure you are on
instance foo or goo.
cfobject action=create type=java
I actually have session sharing off now, using sticky sessions. Yeah, it was
pretty kewl when I had it enabled. The issue we had was with the charting
engine and we had one app where the developer stored CFCs in the session
scope. The charting engine will not work with session sharing on and
I actually have session sharing off now, using sticky sessions. Yeah, it was
pretty kewl when I had it enabled. The issue we had was with the charting
engine and we had one app where the developer stored CFCs in the session
scope. The charting engine will not work with session sharing on and
* No CFCs in session scope (Has this changed in CFMX7? Guess
I can look it up.)
I'm pretty sure that hasn't changed.
Dave Watts, CTO, Fig Leaf Software
http://www.figleaf.com/
Fig Leaf Software provides the highest caliber vendor-authorized
instruction at our training centers in Washington
Anyway, say I've got CFMX clustered--one instance on Machine
A and another on Machine B:
If something hoses up CF on Machine B (an endless loop, for
instance), I don't want Machine A's cluster to keep trying to
serve up my CF app from Machine B.
If this is a JRun cluster, it should stop
Is it possible to share sessions without clustering?
(I'd like to be able to know that when I'm hitting Server A, that the
application I'm seeing is coming from Server A. This is for hardware
load-balancing and fail-over reasons.)
Thanks,
Jamie
You could use client variables (but I don't recommend it).
You could put your session data into a database. It's about the same,
but gives you more control and doesn't require clustering (or the
expensive CF Enterprise ed.)
You could roll your own clustering type of software... serializing and
Thanks for the reply, Nathan.
Let's say that I've considered those options (and I actually have), but decided
to try J2EE session sharing.
The session sharing part works great, however, I *think* I don't want the
application sharing bit, but maybe someone can talk me out of my reservations
Let me know if I'm getting it. ;)
I think the problem boils down to your HW LB device being smart enough.
I am guessing you have layer 4 monitoring on your servers, aka, If I
can ping it, it's alive. That's not usually good enough. Use layer 7
monitoring, make a test CF page that says ok or
Let me know if I'm getting it. ;)
Yes, and thanks a ton for sticking with this thread...
I think the problem boils down to your HW LB device being smart enough.
I am guessing you have layer 4 monitoring on your servers, aka, If I
can ping it, it's alive. That's not usually good enough. Use
Ok, I see, I think I get it. :D
You've got A and B. A is the primary and it is offloading some
processing to B through your JRun cluster. In this case, you don't need
a hardware load balancer, as your site is on one real server.
Right now, server B doesn't have/need/use an HTTP server. Run a
Ok, I see, I think I get it. :D
You've got A and B. A is the primary and it is offloading some
processing to B through your JRun cluster. In this case, you don't need
a hardware load balancer, as your site is on one real server.
Right now, server B doesn't have/need/use an HTTP server. Run a
Anyway, say I've got CFMX clustered--one instance on Machine
A and another on Machine B:
If something hoses up CF on Machine B (an endless loop, for
instance), I don't want Machine A's cluster to keep trying to
serve up my CF app from Machine B.
If this is a JRun cluster, it should stop
17 matches
Mail list logo