Re: [mapguide-users] MapGuid clustering

2014-08-11 Thread Berg, James
I am also interested in the answer to this question. The scaling of my 
application hasn’t reached that point yet, but the only scaling solution I’ve 
come up with so far is to have multiple servers each with their own instance of 
MGOS running a portion of the maps we serve.

The strategy is de-centralized and doesn’t implement any failover capability 
and each server can’t share resources when one is stressed. This isn’t an ideal 
solution.
___
mapguide-users mailing list
mapguide-users@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/mapguide-users

Re: [mapguide-users] MapGuid clustering

2014-08-11 Thread Mark Volz (LeapCAD)
It really depends on how you are using Mapguide as to which option would be the 
best. I’ll outline the 2 methods I have used in the past but of course there 
are multiple ways of dealing with the issue.

 

Option 1 – out of the box approach with a single, outward facing, server and a 
number of slave map rendering servers.

This is what you get with MapGuide, with no customization, just a bit of 
configuration. It’s been awhile since I have done this but from memory the 
process is:

1)  Install all of MapGuide (server and web components) as per normal on 
server A.

2)  On server B install only the MapGuide server component and when asked 
either enter the static IP address of Server A or its name if using dynamic 
IP’s.

3)  On server C install only the MapGuide server component and when asked 
either enter the static IP address of Server A or its name if using dynamic 
IP’s.

 

I think I am missing one step but that is basically it. Server A will then 
handle all the web service calls and server B, C, etc will handle the map 
generation. Just make certain that you have the same FDO providers on all the 
servers. There used to be some examples of this process on the MapGuide Wiki 
but I couldn’t find them.

 

Option 2 – Using a load balancer, to control multiple MapGuide servers.

This option allows for failover and I’ve find it the best option if you are 
using a tiles. There are a lot of load balancers out there so I’m not going to 
name any products, it all depends on your budget and preferred OS. The basic 
process is this;

1)  Install all of MapGuide (server and web components) as per normal on 
server A.

2)  Install all of MapGuide (server and web components) as per normal on 
server B.

3)  Install all of MapGuide (server and web components) as per normal on 
server C.

4)  On Server A create your data connections, layers, maps, tiles, custom 
code, etc, etc.

5)  Create a package on server A and copy it and customizations to servers 
B  C

6)  Import the package on servers B  C, and tiles, code, etc.

7)  Install the load balancer on server D and configure a common base URL 
which will point to the same shared folder on Servers A, B,  C i.e. 
http://serverD/maps will point to http://serverA/mapguide; 
http://serverB/mapguide, http://serverC/mapguide

8)  The load balancer will then automatically switch the user between A, B, 
 C depending on server’s status and current load

 

The problem with both of the above options is a potential for file lock issues 
if you are using file based data sources. This can be resolved by having a 
central file share which updates local copies of the files on each of MapGuide 
installations.

 

Hope that helps.

 

Mark

___
mapguide-users mailing list
mapguide-users@lists.osgeo.org
http://lists.osgeo.org/mailman/listinfo/mapguide-users