This time I wanted to build up a Cluster for GeoServer. Suppose to have many requests to serve and tasks processed are really high cpu intensive, what you would initially think as first option is to put 2 GeoServers and with a simple proxy balancer switch the traffic 50-50 to each of the 2 nodes. That’s correct, but since the 2 instances have their own “installation” directory they could theoretically provide different data, styles, shapefiles, users and so on.
As explained in many documents or books (e.g “Geoserver. Beginner’s Guide“) the concurrence is not a problem for data who reside on DBs, but what happens for other useful data? (shapefiles and styles in particular). You should pay twice the space, by duplicating them, and moreover you should keep care of syncing all of them!
What was suggested, even from the link of above, was to think about a clustered file system. So how could you do it? Trash away all the documentation about ricci, lucci and so on these stuffs are outdated.
Since last month I’m studying a nice topic like the “High Availability” on Linux OS. If you have a SLES 11 and a license for High Availability Extension add-on, this guide could be really helpful for a good understanding. The lack of official and well documented procedures it’s really common for these graceless topics, but this one with its 495 pages looks awesome.
It’s a really comprehensive manual accompanied with simple examples and many pictures & screenshots. If you want to get familiar with words like Corosync, OpenAIS, STONITH and other, please give it a try. In case you have previous versions like the HA Extension SP1, most of commands seem to work perfectly. From the release notes of SP2 you can see that big changes to components haven’t been done, and that’s what care most 😉 (especially for enterprises environments)
Thanks SuSE for the great job, that’s why I like this distro for Enterprise purpose! Here the link to the page for the .pdf file.
[Update: the guide has been updated on June 26th 2013!!}