How to Implement a Cluster of Clusters

Atiq AdamjeeSenior ArchitectNovell, Inc.

Brad RuppSoftware EngineerNovell, Inc.

© March 9, 2004 Novell Inc.2

one Net: Information without boundaries…where the right people are connected with the right information at the right time to make the right decisions.

The one Net vision

Novell exteNd™

Novell Nsure™

Novell Nterprise™

Novell NgageSM

:

:

:

:

© March 9, 2004 Novell Inc.3

The one Net vision

Novell Nterprise is an innovative family of products which gives you the power to enable and manage the constant interaction of people with your business systems — regardless of who they are or where they are.

Novell Nterprise™

Novell exteNd™

Novell Nsure™

Novell Nterprise™

Novell NgageSM

:

:

:

:

© March 4, 2004 Novell Inc.4

Agenda

Introductions

Assumptions

Components of a Disaster Recovery Plan

Data Replication

Factors Affecting Data Replication

Distance Consideration

Latency Consideration

Bandwidth Consideration

Cluster of Clusters Implementation

Q & A

© March 4, 2004 Novell Inc.5

Introductions

Atiq AdamjeeSenior Architect, Novell Consulting

atiq.adamjee@novell.com

Brad RuppSoftware Engineer, Novell Inc.

brad.rupp@novell.com

© March 4, 2004 Novell Inc.6

Assumptions

Knowledge of Novell Cluster Services™ (NCS)

Have already implemented NCS or are in the process of implenting NCS

Need to implement a disaster recovery plan

This is not a full disaster recovery plan – but a large component of it

© March 4, 2004 Novell Inc.7

Components of a Disaster Recovery Plan

A complete disaster recovery plan must include more than just “servers”

People

Facilities

Information Technology

Much more...

© March 4, 2004 Novell Inc.8

Data Replication

Data may be replicated in many different ways

Host-based

Disk-based

Asynchronous

Synchronous

© March 4, 2004 Novell Inc.9

Host-based Replication

FCWDM

SONETATMIP

Etc.

Data Center Recovery Site

© March 4, 2004 Novell Inc.10

Disk-based Replication

FCWDM

SONETATMIP

Etc.

Data Center Recovery Site

© March 4, 2004 Novell Inc.11

Asynchronous Replication

replic

ate

write

write

Latency of I/O to remote disk does not slow down host I/O

Data on the remote disk is only as current as the last replication session

© March 4, 2004 Novell Inc.12

Synchronous Replication (Mirroring)

Latency of I/O to remote disk will slow down host I/OData on the remote disk is always current

write

write

write

2ms

15ms

© March 4, 2004 Novell Inc.13

Factors Affecting Data Replication

Distance

Latency

Bandwidth

2

1

3

© March 4, 2004 Novell Inc.14

Distance Consideration

Distance Consideration

Within a campus

Under 10 km

Under 100 km

Over 100 km

© March 4, 2004 Novell Inc.15

Distance Consideration

Within a campus - use Ethernet or fibre channel

Under 10 km – consider fibre channel

Under 100 km – fiber, WDM, FC/IP, ATM, SONET, iSCSI, …

Over 100 km – iSCSI, fiber, WDM, FC/IP, ATM …

As distance increases so does the cost of the connection

iSCSI has changed this by allowing SCSI data to be routed over TCP/IP.

© March 4, 2004 Novell Inc.16

Latency Consideration

Shorter round trip delay = more frequent transmission

Impacted by overhead, i.e. how much of each packet is really data

With longer distances, latency impacts performance more than bandwidth

© March 4, 2004 Novell Inc.17

Bandwith Consideration

What is the amount of daily change in the data?

How much time do you have to replicate the data?

What is the round trip delay between this site and the distant replica?

What is the overhead for the selected transport?

© March 4, 2004 Novell Inc.18

Cluster of Clusters

Powered by NetWare and Novell Cluster Services

Manually fails over network services from one cluster to another geographically, separate cluster

Provides high availability to your mission critical resources in the event of a disaster

SBD and heartbeat issues go away

© March 4, 2004 Novell Inc.19

Cluster of Clusters Implementation Overview

Real World Implementation of a Cluster of Clusters

Six servers at the primary site

Two servers at the disaster recovery (DR) site

EMC CLARiiON SAN at both sites

Both SANs mirrored over an OC3 link

© March 4, 2004 Novell Inc.20

Cluster of Clusters Implementation – Phase I

Unmask the LUNs on the SAN so all servers at the site can see all shared devices

If using existing NSS pools, mark them as shareable for clustering

Install NCS

Activate the pools and volumes

Create cluster volume objects and virtual servers

Configure resources, load/unload scripts, failover priorities, etc.

Bring resources online

Test failover between servers

© March 4, 2004 Novell Inc.21

Cluster of Clusters Implementation – Phase II

Down the cluster at the primary site

Stop SAN mirroring

Mask (hide) the LUNs from the servers at the primary site

Remove the LUN masking (make the LUNs visible) from the servers at the DR site

Repeat the steps in Phase I to set up the DR cluster

© March 4, 2004 Novell Inc.22

Cluster of Clusters Implementation – Phase III

Down the cluster at the DR site

Mask (hide) the LUNs from the servers at the DR site

Start SAN mirroring

Remove the LUN masking (make the LUNs visible) from the servers at the primary site

Start the primary cluster

© March 4, 2004 Novell Inc.23

Cluster of Clusters Implementation – Phase IV

Prepare for disaster recovery

Login Scripts – create DR scripts ahead of time and comment them out or use tools like the NDS Login Script editor available from Cool Solutions

Printing – set up printers at the DR site as printers are location-based entities

eDirectory replicas

Other services, e.g. DHCP, DNS, SLP, etc.

© March 4, 2004 Novell Inc.24

Cluster of Clusters Implementation – Phase V

Test Cluster Failover

Down the cluster at the primary site

Stop the SAN mirroring

Mask (hide) the LUNs from the servers at the primary site

Remove the LUN masking (make the LUNs visible) from the servers at the DR site

Enable the cluster (and mount the volumes) at the DR site

Make necessary changes to the login scripts

Restore other services, e.g. printing, DHCP, DNS, SLP, etc.

© March 4, 2004 Novell Inc.25

Cluster of Clusters Implementation – Phase VI

Test Cluster Failback

Down the cluster at the DR site

Mask (hide) the LUNs from the servers at the DR site

Remove the LUN masking (make the LUNs visible) from the servers at the primary site

Enable the cluster (and mount the volumes) at the primary site

Make necessary changes to the login scripts

Restore other services, e.g. printing, DHCP, DNS, SLP, etc.

Question and Answer

© March 4, 2004 Novell Inc.28

General DisclaimerThis document is not to be construed as a promise by any participating company to develop, deliver, or market a product. Novell, Inc., makes no representations or warranties with respect to the contents of this document, and specifically disclaims any express or implied warranties of merchantability or fitness for any particular purpose. Further, Novell, Inc., reserves the right to revise this document and to make changes to its content, at any time, without obligation to notify any person or entity of such revisions or changes. All Novell marks referenced in this presentation are trademarks or registered trademarks of Novell, Inc. in the United States and other countries. All third-party trademarks are the property of their respective owners.

No part of this work may be practiced, performed, copied, distributed, revised, modified, translated, abridged, condensed, expanded, collected, or adapted without the prior written consent of Novell, Inc. Any use or exploitation of this work without authorization could subject the perpetrator to criminal and civil liability.