What is MC Service Guard?

[Verónica Muñoz Segovia]

Hello!

Here I am again, I'm looking for a high availability on my system, and I know a Cluster can help me to have this redundancy in case of one of the servers goes down.

I don't know this product, can anybody tell me something about it?

How expensive the product is?

Is there any other altertative to can have redundancy?

My idea is to have two servers for production and one spare server and if one of the production server goes down the spare server take the personality of the down server.

Is there any possibility to can do this manually without Service Guard? What are the advantages and disadvantages?

The three servers have HP-UX 10.20 and are two K370 and one K580 with SCSI cards.

Thanks in advance

Veronica

[Santosh Nair_1]

MC/ServiceGuard is HP's high availability solution. It makes sure that the application is protected from hardware/software failures by moving the application from a "bad" machine (node) to a "good" machine (node). Where a bad machine is one where the application has problems running because of lack of resources or hardware problem and the good machine one where the application has no problems running.

For more details see the following URL:

http://www.hp.com/products1/unix/highavailability/ar/mcserviceguard/index.html

-Santosh

[Sanjay_6]

Hi Veronica,

Dowload the pdf below for a helpful reading

http://docs.hp.com/hpux/onlinedocs/ha/hpworl98.pdf

and the SG FAQ

http://docs.hp.com/hpux/onlinedocs/ha/haFAQindex2.html

Hope this helps.

thanks

[A. Clay Stephenson]

Hi Veronica:

This link will explain the product:

http://www.hp.com/products1/unix/highavailability/ar/mcserviceguard/index.html

Basically the idea is somethin like this:

HostA ----- Data + Package IP Address ---- HostB

In this (a two node cluster), the package (e.g. Oracle) has a volume group(s). The volume group(s) is active and mounted on only one host but if that host fails the volume group(s) is activated and mounted on the other host. The other key is that the clients do not connect directly with HostA or HostB but with a package IP address that moves with the package.

You can have many packages on a cluster and in our example PkgA could normally run on HostA and PkgB & PkgC could normally run on Host B.
If any node fails, all the packages can shift to the remaining node. There is no need to have a box sitting around just waiting for a package to fail - if you have more than 1 package.

If this is truly a high availability application then I would definitely go with MC/ServiceGuard.

The latest list prices I have are $3900 for the K3's and $11,900 for the K5xx. When you start looking at MC/SG the cost off the software is not the total picture. You have to look at redundant network cards, redundant network switches, redundant HVAC, rudundant data paths and mirroring, backup generator to supplement your UPS, ... .

Food for thought, Clay

[Santosh Nair_1]

Opps forgot the rest of the question...

As for pricing, I believe it's in the 10K/machine region...but consult your local HP rep for better pricing details.

The good thing about SG versus is that it doesn't have the concept of active and passive machines...so you could have applications A and B running on machine 1 and applications C and D running on machine 2. Application A and B can failover to machine 1 and C and D to machine 2. So you can make better use of your hardware since you don't have a machine sitting idle just waiting for a crash.

As for alternative solutions...I suppose you could duplicate everything that SG does with scripts, but there would be some intervention required, i.e. to determine when to fail the package over.

Another alternative is MON and heartbeat from the Linux HA project:

http://www.linux-ha.org

-Santosh

[John Poff]

Hello,

We use MC ServiceGuard. We have clusters here and at remote sites. The advantages of using it is that once you have it configured it is very good and reliable. I worked with one of our remote sites yesterday that had a system failure, and the second machine took over and had them up and running five minutes later. This application was at a remote distribution center, so we could have had dozens of people standing around doing nothing while the system was down. As for the cost, you have to weigh that against the cost of being down. In our case, just being down a couple of hours would cost us more than buying and implementing ServiceGuard, so it is a bargain for us.

The disadvantages are that you will have to learn ServiceGuard. The best way to do that is to take the classes from HP; they are excellent.

I've never tried doing something like that manually, but there have been some discussions here in other threads about how it might be donw. It didn't sound very easy, and the people that knew how to do it manually recommeneded using ServiceGuard. My guess is that in the time it would take you to figure out how to do it manually, you could already have it done using ServiceGuard.

Your hardware configuration and O/S version is the same as what we have in our remote sites, so you should be fine. You will need both machines to have access to shared disk so that each machine can mount the volume groups as needed.

JP

[Sanjay_6]

Hi Veronica,

Have a look at the thread below for pricing info. Do remeber that if you take MC/SG, you should take Mirrordisk/UX too.

http://www.software.hp.com/cgi-bin/swdepot_parser.cgi/cgi/displayProducts.pl?group_type=category&group_name=HA

Hope this helps.

thanks

[James R. Ferguson]

Hi Veronica:

In my opinion, another good starting point for concepts, at least, is Peter Weygant's "Clusters for High Availability, A Primer of hp Solutions" (Copyright 2001 ISBN: 0-13-089355-2). See:

http://www.hp.com/hpbooks/prentice/ptr_0130893552.html

Regards!

...JRF...