Solaris

Good day, everyone, and thanks first off for reading my question.

So, I have been Googling and reading oracle documentation for the past couple weeks, and I am just getting more and more confused as to what I need to do, and I would really appreciate some guidance or at least pointing me to what specifically I should be reading.

Situation: We have several 10+ year old Solaris systems, running either Solaris 8 or 10, that hold several mission-critical systems for business operation. These systems are rapidly dying and need to be migrated to a newer system. They are old systems that require very few resources to run, but are critical nonetheless.

So far: We have purchased four T5220 SPARC servers, with 8 cores and 64GB of RAM each. They have arrived, and I am ready to start building the "solution".

The intent: To be able to take system images of these old systems, virtualize them, and have them run on some type of high-availability cluster solution that will keep them operational, even if one (or possibly two) of the four physical T5220 servers failed.

Question: How should I do this? It seemed like a simple idea to start out with. My initial thoughts were to install Solaris 11 on each of the four servers, cluster them together with Oracle cluster, run a Hypervisor of some type as a cluster application (Oracle VM?) on top of the physical cluster, and then drop all of the virtual machines on top of that hypervisor. That way, the virtual machines should run uninterrupted if a cluster node fails.

Is this feasible? Possible? What is the right terminology here so I know what to be researching? I didn't think this would be so hard, but I can't even find a good white paper/document to read - they all seem to be focused to something other than what I'm looking for.

Any suggestions for how I should accomplish this or where I can go to find the information I'm looking for?

Thanks in advance for all your help.

-Lyxix

LDOM (Oracle VM server for SPARC), with HA LDOM solaris cluster between physical nodes, it will cost you since SC is 'per core'.

Trouble is : If a node fails for some reason, cluster will only boot the LDOM on the next host (couple of minutes of downtime, but downtime).

You can migrate old machines to future LDOMs using flar archive or similar method.

Option is using LDOMs and running Solaris cluster between virtual machines. Depending on your resource (application) you will be able to configure active / active or active / passive.

If a node dies, other node and its ldom with SC will takeover, while you can just import the rest of the machines (ldoms) to working node (what SC HA LDOM does, but automatically, in above example).

Be sure to have auto-boot?=false for all ldom's which are on shared storage.
You don't want to boot root zpool on two nodes at once

You might also use bare metal with branded zones (doubtful solaris 8 will work that way) and Solaris cluster with HA ZONE or Solaris cluster zone cluster for high availability.
Or just importing the zone manually if the node dies, since zone rpool will not import on another node if configured properly (it will give error that it is active on another node).

Hope that helps
Regards
Peasant.

Peasant:

Thanks so much for your help, and what a huge help it was! I read what you wrote, researched a bit more on LDOMs (that I now know is synonymous with Oracle VM Server, thanks to you), and read an article that someone suggested to me on another site:

ht tps :// blogs.oracle.com/orasysat/entry/zones_clusters_clustering_zones_zoneclusters

This is what it looks like I need to learn about and implement. How does this sound for an extremely broad overview:

Step 1) Cluster the 4 physical servers together (figure out what type of cluster I need, and use Oracle Cluster 4.x on top of Solaris 11.1 or maybe 11.2 beta?)

Step 2) Create Zoneclusters, probably one for each type of OS I need to run (a Sol 8 ZC, and a Sol 10 ZC, and then possibly a Sol 11 ZC to try to migrate the systems up if possible...)

Step 3) Create Logical Domains (LDOMs) for each of my old systems, and put them on the zone clusters, using Oracle VM Server.

In this setup, if a node fails, the virtual system will go down, but I should be able to configure it to start up another node automatically with minimal downtime for the virtual systems?

I realize "I should be" is reaching, since I have virtually zilch cluster experience and limited Solaris experience, but let's rephrase that to "it's possible to". I realize I have a lot of work ahead of me, but thankfully I have no strict time limits to get this implemented and can work one step at a time.

As far as cost goes, the one good thing is that, per the Oracle license agreement, you don't need to pay for SC while it's in the development stages. Directly from the OTN License Agreement for Oracle Solaris Cluster:

"You may not:

- use the Programs for your own internal business purposes (other than developing, testing, prototyping and demonstrating your applications)"

Only after I get the thing actually working and implemented will we have to shell out the big bucks. Presuming I do get it working, I will have achieved the "proof of concept" with these cheaper T5220 servers and the company will then be willing to buy some nice T4's or better to go along with the pricey software.

Thanks so much for your help - does this all sound correct so far?

-Lyxix

Best scenario you would like to use is (and cheapest) :

Install Solaris 11.1 (do not install beta version), it will come with Oracle VM server for sparc on all physical nodes since you currently have multiple version of Solaris in production.

Check to have everything the same version on all physical hosts, this is important (firmware, Os version etc.)

Configure it, and create as much ldoms as you need and migrate your Solaris installations on it.
Oracle VM Server for SPARC Physical-to-Virtual Conversion Tool - Oracle VM Server for SPARC 3.1 Administration Guide

Requirement is to have FC or similar tech (NFS, ISCSI), so all the physical hosts see all the network disks and are logically configured with same names on all physical hosts.

This will enable you to live migrate a ldom from host to host with no downtime or import tje configuration of ldom and boot it on another host in case of failure (you will need to backup ldom xml configuration from all physical hosts on external location).

You can, of course, have local disks ldoms, but those are prone to single host failure, but such configuration can be used if you have SC between two ldoms on two physical hosts.

In the above example, you will not be using SC, for SC scenario is much alike but then clusterware is doing all the migrations and booting in case of failure or switching.

Thanks for all the help, Peasant. That's exactly what I was hoping for. I'll do just that.

-Lyxix