Operating System - HP-UX
1748270 Members
3760 Online
108760 Solutions
New Discussion юеВ

Re: Hot Standy w/ ServiceGuard

 
SOLVED
Go to solution
Anthony Baldwin
Occasional Advisor

Hot Standy w/ ServiceGuard

Hello. Can someone explain to me the terminology of hot standby? I have been charged w/ the task of configuring a MC/SG cluster, implementing "hot standby.... For example, the customer wants to have 2 node cluster, remote from one another. The second node is a standby that will become the primary in the event of a disaster."

The customer wants to have Oracle database among its software packages that MC/SG monitors. Is this how we want to use MC/SG? I had the pleasure of taking the MC/SG course, but I gathered that you would want to use ServiceGuard in conjunction with mirroring, where the nodes in the cluster (at least 3 nodes) have access to the same data disks (preferrably, an external disk device).

What the customer sounds like they want is a disaster/recovery setup (if one box fails the other can be used)... The question is, do you want to use ServiceGuard in this way and is this concidered to be "hot standby"? Please advise. Thanks.

A.J.B.
Wisdom is the principal thing; therefore get wisdom: and with all thy getting get understanding.
4 REPLIES 4
Pramod_4
Trusted Contributor
Solution

Re: Hot Standy w/ ServiceGuard

Hi,

There are two type of high available solutions:

*) SFT clusters ? system fault tolerant clusters, for eg:- Novel NetWare SFT
*) High available cluster , For Eg- HP MC/SG, Sun Cluster etc.

In SFT both the systems are mirrored, to be more precise, at any point of time memory of the cluster members would be in sync. So in case of the failure of the primary server, Not only the applications but also the last cursor position of the user screen would be intact., Typically user would never come to know of the failure.

But in High available clusters such as Sun or Hp clusters, in case of oracle like RDBS, up to the last committed transactions can be retrieved. I mean to say what ever was not written to the disk by the primary server would be lost in the event of a failure.


As far as MC/SG is concerned it does not have any protection / monitoring for disk failure. So you need to use either Mirror UX or RAID 0/1 for data redundancy. As you are thinking to make it as DR capable setup, consider to keep the different copies of the mirror geographically apart ( you may have to think of using fiber link).

HP offers standard MC/SG toolkit for Oracle, so the integration is going to be easy.

Finally MC/SG in conjunction with Disk mirroring ( either hardware of software ) you can keep your application up and running from the single point of failures such as

Network interface, Disk, application itself on a given box, or the CPU box itself.

Regards,

Pramod









James R. Ferguson
Acclaimed Contributor

Re: Hot Standy w/ ServiceGuard

Anthony:

With MC/ServiceGuard, various configurations are possible. An "active/standby" configuration is one in which a standby server is configured to take over after the failure of another. In this configuration the backup node may be idle or it may be running a less important function.

An "active/active" configuration is one in which several nodes run critical apps while some serve as backups for others while still running their own applications.

An excellent overview of high-availability is an HP professional book "Clusters for High Availability: A Primer of HP-UX Solutions" by Peter S. Weygant (ISBN 0-13-494758-4).

You might also want to look at the "Designing Disaster Tolerant MC/ServiceGuard Clusters" manual:

http://docs.hp.com/hpux/onlinedocs/B7660-90001/B7660-90001.html

...JRF...
James R. Ferguson
Acclaimed Contributor

Re: Hot Standy w/ ServiceGuard

There are other responses in the duplicate post of this thread.
alexander niessen
Occasional Visitor

Re: Hot Standy w/ ServiceGuard

Hi Anthony,

MC/SG is a product which allows you to build High Available Cluster configuration, by creating redundancy in Hardware. MC/SG monitors hardware and software resources and can handle so called SPOFs (single points of failures). A SG cluster contains minimum out of 2 nodes which should be configured in a way to take over each others applications. An application, like Oracle, is configured in a so called package. This package has to be configured to run on both nodes of the cluster.

A hot standby configuration means that one node of the cluster is in a idle state and only becomes active when for any reason the active cluster node fails. Hot standby, also means, that the standby node takes over control automaticaly and usualy no user intervention is needed to switch the package on this node. By the way, when speaking about MC/SG, we always talking about hot-standby.

It seems to me that the implementation your customer wants is not clear to you. So I will give you some questions you have to solve, before going on with this implementation:

1. Wants your customer a physical seperation of his datacenter?
If he wants this, you have to implement a so called Campus or Metro Cluster

2. Is the needed Hardware and Software already there, or do you have to deal with this things too.

3. Is there already in installed implementation of this Oracle DB, should you migrate this, or is this a new installation?

4. Does the customer already run MC/SG in his IT?

To get a more detailed information on MC/SG please visit:

http://docs.hp.com/hpux/ha/index.html

To understand HA concepts the whitepaper recommended in the above reply is a good start.