Service Guard Setup

[Romanek]

Hello

I've just started with a company with several HP 9000 rp3440 servers and a MSA 1000 SAN. The company wants Service Guard setup on two of the rp3440s so that their custom application (baically a database system) can fail-over between the two box in case of problems. I have never setup Service Guard before and the company wants to go live in 6 weeks.

My question is: How hard is Service Guard to setup and get working well? ...say on a scale from 1 to 10? Can anyone point me to any good how-tos out there? Any words of wisdom?

Thanks!
Greg

[melvyn burnard]

1. Do a search in the forums for a lot of threads asking similar questions.

2. Read the Managing Serviceguard manual for the version you wish to use at http://docs.hp.com/en/ha

3. Enrol in the HP Customer Education class for Serviceguard

4. Consider using a Consultant to get the cluster installed and running, and sitting in on the work for your own education, but make sure you do the reading first.

On a scale of 1 to 10, not easy to answer, but could be as low as 1, as high as 9
It all depends on you existing knowledge, and experience

[A. Clay Stephenson]

http://docs.hp.com/en/ha.html#Serviceguard

To configure MC/SG and get it up and running is probably a 4; however, to truly achieve a high-availability system so that unplanned downtime actually approaches zero is a 9 or 10.

Generally, MC/SG is easy to install and configure. I recently configured an MC/SG cluster with 3 packages in under 2 hours although I have had just a bit of experience. What I'm not telling you is that the prerequisites (redundant disk, multiple SCSI paths, multiple independent networks, redundant power, procedures, ... ) took much longer than 2 hours or 2 weeks. Before you even begin to consider MC/SG, you need to eliminate essentially all single points of failure.

In addition you need to be able to LVM and/or VxVM in your sleep before tackling MC/SG.

Surprisingly, I have never had an actual failover so that MC/SG had to come to the rescue. You see, by the time you've done all the prerequisite stuff (Redundant power -- and I mean UPS + generator, redundant HVAC, redundant networks, redundant disk arrays, ...), MC/SG very seldom comes into play itself. I'm not saying don't bother with MC/SG; quite the contrary. MC/SG imposes a level of discipline on you and if you are not willing to commit to the underlying infrastructure then the gains from MC/SG itself are quite limited.

If you already know LVM and/or VxVM then I would strongly suggest that you take the MC/SG class. You will be able to yank on cables and pull disks out by their roots on someone else's equipment and learn about MC/SG at the same time.

[Prashanth.D.S]

Hi,

I agree to both version of replies..try browsing the below link you would get more information on SG.

Since the SG is used to maintain High Availability its recommended to attend the course for SG before you start with implimentation.

http://docs.hp.com/en/B3936-90100/index.html

Best Regards,
Prashanth

[Thomas J. Harrold]

Word of wisdom: Keep it simple at first, add complexities later.

Test your failover to make sure that it all works as you expect.

If you have to make ANY change, re-test. The most common cause of cluster failures is admins making changes, and NOT testing. (typos, etc...)

I have a SG checklist that I can dig up for you, if you are ready to start playing with MC/SG.

-tjh

[Romanek]

Thanks to everyone for the replies. They were helpful!

Greg

[Romanek]

Closing thread. I wanted to get some feedback on the level of difficulty of Service Guard to help me proceed. Thanks!