I've been intrigued by Solaris Zones since I first heard about the project several years back. Thus, I was excited to be given the chance to help lead the team responsible for ensuring that Sun Cluster 3.2 take full advantage of this great new virtualization technology.

Now that Sun Cluster 3.2 is in beta, I'm pleased to report that we've done an outstanding job (if I do say so myself) of integrating Sun Cluster with Solaris Zones. Sun Cluster now supports zones as full-fledged “virtual hosts” of cluster services. Using zones with Sun Cluster is easy because the Sun Cluster resource management interfaces for using zones and physical nodes are identical. Practically, that means you can just specify a zone instead of a physical node in a Resource Group nodelist and the cluster will run the application inside that zone instead of in the global zone, with most of the isolation and security benefits thereof. And the best part is that the Sun Cluster agents (data services) don't need to be modified at all to run in the non-global zones. For example, the same HA-Apache agent that runs in global zones works perfectly in non-global zones.

So why should you care? Here are a couple reasons why you might want to use zones with Sun Cluster:

• Testing: If you have limited hardware, or just don't want to bother configuring a full test cluster, you can prototype on a single physical machine with multiple zones. There are a few caveats (the Shared Address logical IP address isn't supported across multiple zones on the same physical machine), but it should generally work fine.

• Deployment Isolation: Suppose you want to run several unrelated services on a single cluster. You can set up zones and run each service in a different set of zones across the physical nodes. In this way you can give yourself several “virtual clusters” on one set of physical nodes. Using the new Service Level Resource Management features of Sun Cluster 3.2 you can assign different CPU shares to each service. And because of zone security, you can even safely delegate administration for each virtual cluster to different administrators.

Now I'll provide a small example of how easy it is to run the HA-Apache service in non-global zones. This example assumes you've already configured apache properly in each non-global zone by providing an http.conf file, etc. Once you've done that, in order to add high availability, simply run the following cluster commands:

1. Register the resource type
   
# clrt register SUNW.apache

   
2. Create the resource group. Note that the nodelist specifies three non-global zones using : syntax. The physical nodes are named ptriped1, ptriped2, and ptriped3. The non-global zones are named lztriped on each node.
   
# clrg create -p nodelist=ptriped1:lztriped,ptriped2:lztriped,ptriped3:lztriped apache-rg

   
3. Add the LogicalHostname IP address (named triped-2 on my system).
   
# clreslogicalhostname create -g apache-rg triped-2

   
4. Add the apache resource:
   
# clrs create -g apache-rg -t apache -p bin_dir=/usr/apache2/bin/ -p port_list=80/tcp apache-res

   
5. Bring the service online.
   
# clrg online -M apache-rg

   
6. Verify the status
   
# clrg status

=== Cluster Resource Groups ===

   Group Name Node Name Suspended Status
   ---------- --------- --------- ------
   apache-rg ptriped1:lztriped No Online
             ptriped2:lztriped No Offline
             ptriped3:lztriped No Offline

   
# clrs status

=== Cluster Resources ===

Resource Name Node Name State Status Message
------------- --------- ----- --------------
triped-2 ptriped1:lztriped Online Online - LogicalHostname online.
         ptriped2:lztriped Offline Offline
         ptriped3:lztriped Offline Offline

apache-res ptriped1:lztriped Online Online - Service is online.
           ptriped2:lztriped Offline Offline
           ptriped3:lztriped Offline Offline


   And in the zone ptriped1:lztriped --
   
# ps -ef | grep httpd
webservd 3651 3646 0 17:17:36 ? 0:00 /usr/apache2/bin/httpd -k start
webservd 3650 3646 0 17:17:36 ? 0:00 /usr/apache2/bin/httpd -k start
webservd 3649 3646 0 17:17:36 ? 0:00 /usr/apache2/bin/httpd -k start
webservd 3648 3646 0 17:17:36 ? 0:00 /usr/apache2/bin/httpd -k start
webservd 3647 3646 0 17:17:36 ? 0:00 /usr/apache2/bin/httpd -k start
root 3661 3407 0 17:18:06 pts/2 0:00 grep httpd
root 3646 2393 0 17:17:35 ? 0:00 /usr/apache2/bin/httpd -k start


That's it! I hope you try out this great new feature when you get your hands on Sun Cluster 3.2!

Nick Solter
Sun Cluster engineering