• No results found

Managing Live Migrations

N/A
N/A
Protected

Academic year: 2021

Share "Managing Live Migrations"

Copied!
5
0
0

Loading.... (view fulltext now)

Full text

(1)

D

ata center managers must perform migrations and upgrades while incurring as little downtime as pos-sible for mission-critical applications. IT departments migrating from Novell NetWare—which is being phased out—to Novell SUSE Linux Enterprise Server face such a challenge. The Novell Open Enterprise Server (OES) platform supports both NetWare and Linux, provid-ing a bridge between these operatprovid-ing systems and an environment for migration. As OES NetWare systems are gradually replaced with OES Linux systems, Novell Cluster Services can help minimize or eliminate down-time during migration by allowing administrators to build heterogeneous server clusters comprising both NetWare and Linux servers. This article describes steps for configuring Linux nodes and adding them to existing NetWare clusters as well as considerations for configuring heterogeneous clusters.

Using Novell Cluster Services

Novell Cluster Services is a multi-node active/active server-clustering tool that can help improve the avail-ability, load balancing, and manageability of critical

network resources, including data, applications, and ser-vices. Novell Cluster Services 1.8, enabled for the Novell eDirectory™ directory service, is the OES version for

NetWare 6.5 and SUSE Linux Enterprise Server 9. Novell Cluster Services allows failover, failback, and migration (load balancing) of individually managed cluster resources. It supports up to 32 nodes, and any node in the cluster can restart resources (applications, services, IP addresses, and volumes) from a failed server in the cluster. All nodes in a given cluster must be configured with a static IP address, and they must reside on the same IP subnet and in the same eDirectory tree.

Novell Cluster Services uses the Split Brain Detec-tor (SBD) for heartbeat functionality. The SBD moniDetec-tors server clusters for nodes that lose their Ethernet connec-tion but retain all other funcconnec-tionality, a state known as a split brain. Because this node can no longer reach the other nodes with Global Interprocess Communication sig-nals or heartbeats, it tries to take over all of the cluster’s resources; meanwhile, the other nodes, thinking that the first node has failed, try to take over its resources. The SBD stops the smaller side of this split brain or the side BY CINDY STAP, DEAN OLIVER, AND AHMAD ALI

Managing Live Migrations

from Novell NetWare to SUSE Linux Enterprise Server

For many IT departments, avoiding downtime for mission-critical systems during

migrations and upgrades is a top priority. The Novell Cluster Services

tool can help

manage live cluster migrations from the Novell

®

NetWare

®

OS to Novell SUSE

®

Linux

®

Enterprise Server.

Related Categories: Clustering Dell PowerEdge servers Novell eDirectory Novell NetWare Novell SUSE Linux

Visit www.dell.com/powersolutions for the complete category index.

(2)

not running the master node. In two-node clusters, because both nodes are the same size and both are running the master node, the SBD checks for LAN failure and stops whichever node has lost its network connection. The SBD uses an 8 MB partition on the shared storage accessible to all nodes in the cluster.

Performing a test migration from NetWare to Linux

In October 2005 a team of Dell engineers configured a test cluster environment to illustrate the migration path from NetWare to SUSE Linux Enterprise Server. The test configuration consisted of a three-node Dell™™ PowerEdge™™ server cluster running NetWare 6.5 with

Support Pack 4 and Novell Cluster Services 1.8 (see Figure 1). The cluster was tied to an eDirectory 8.7.x tree. Each server was config-ured with dual Gigabit Ethernet1adapters for LAN connectivity and

QLogic Fibre Channel controllers for storage area network (SAN) con-nectivity. The cluster also included a Dell/EMC CX600 storage array with multiple logical units (LUNs). Novell Storage Services™™pools

were configured on these LUNs as highly available clustered storage resources and provided highly available file services as Common Internet File System (CIFS) resources to cluster clients.

Building the Linux nodes

In the test configuration, the first Linux node was built by retir-ing one of the existretir-ing NetWare nodes from the cluster. First, all of the node’s resources were migrated to the remaining two NetWare nodes. In addition, the node was removed from the NDS tree and disconnected from the SAN. Then SUSE Linux Enterprise Server 9 was installed on this node, and the default OES configura-tion was selected. After the OS installaconfigura-tion, the node was added back onto the NDS tree.

Novell Storage Services and Novell Cluster Services were added in the software selection of the YaST (Yet Another Setup Tool) instal-lation and systems management application. For Linux servers, YaST includes plug-in utilities for both Novell Storage Services and Novell Cluster Services. YaST is the main tool for deployment and configu-ration of Linux nodes in Novell Cluster Services. In OES NetWare, by contrast, cluster installation, administration, and upgrades are performed through the Novell Deployment Manager utility, which runs on a Microsoft® Windows® client.

Configuring Novell Storage Services. To configure Novell Storage Services, administrators should open YaST and select System>NSS. They can then authenticate to the tree by entering the Lightweight Directory Access Protocol (LDAP) server, adminis-trator, and password information. A unique Novell Storage Services administrator object name is created.

Configuring Novell Cluster Services. To configure Novell Cluster Services, administrators should open YaST and select

System>NCS. They can then enter the LDAP server, administra-tor, and password information. On the next screen, they should choose “Existing Cluster” and enter the fully distinguished name (FDN) of the cluster. Because failure can occur at this point if the Linux server cannot find an existing SBD partition on the SAN, best practices recommend verifying the partition with the nssmu util-ity before performing this step. Finally, administrators can use the Novell iManager console to verify that the Linux server node has been added to the cluster, as shown in Figures 2 and 3.

Figure 4 lists some of the packages and required services installed during Novell Cluster Services installation on the Linux server node.

Configuring heterogeneous cluster environments

After Linux nodes have been added to a NetWare cluster environ-ment, Novell Cluster Services pushes down various cluster load and unload scripts, XML files, and .conf files in the /etc/opt/ novell/ncs and/var/opt/novell/ncs directories.

Many of the files contain NetWare-specific commands, which are translated by the ncs-resourced daemon. For example, the NetWare commandcluster cvsbind add vserver 10.0.0.0 is translated to the following Linux command:

1This term does not connote an actual operating speed of 1 Gbps. For high-speed transmission, connection to a Gigabit Ethernet server and network infrastructure is required.

Figure 1. Cluster configuration for test migration from NetWare to Linux

Authentication server Clients

PowerEdge server PowerEdge server PowerEdge server Public network Private network

SAN

Dell/EMC CX600 storage array

(3)

ncpcon bind –ncpservername=vserver --ipaddress=10.0.0.0

This translation process imposes a restriction on mixed-node environments: Because NetWare commands can be translated into Linux commands but not vice versa, cluster resources can be cre-ated only on NetWare nodes. This restriction becomes irrelevant once the migration process is complete and all of the Novell Cluster Services nodes are running Linux.

Administrators also should note the following about mixed-node cluster environments:

Online storage reconfiguration is not supported.

NetWare nodes cannot be added using Novell Deployment Manager.

Resources created on Linux cannot run on NetWare.

Other kernel modules, plug-ins, and paths that would be of interest for administrators are as follows:

/opt/novell/ncs /lib/modules/'uname -r'/ncs/*.ko /usr/include/ncssdk.h /usr/lib/libncssdk.so /proc/ncs /admin/Novell/Cluster

Novell Cluster Services supports fast-mount journaled file sys-tems such as Reiser, ext3, and Novell Storage Services. Linux sup-ports all of these file systems, but NetWare only supsup-ports Novell Storage Services. Therefore, in heterogeneous cluster environments with Novell Cluster Services, only Novell Storage Services can be used as the file system for storage resources.

Creating the SBD partition

For heterogeneous clusters, Novell recommends creating the SBD partition from NetWare, a task that is usually performed during cluster creation. In the test configuration, the SBD partition had already been created from the NetWare node. If necessary, however, the partition can be created using the following commands:

SBD INSTALL (NetWare)

sbdutil -c -d device_name (Linux)

If the partition is created from Linux, administrators should enable the NCS:Shared Disk Flag attribute: In iManager, they should select eDirectory Administration>Modify Object>Cluster Object>Valued Attributes.

Viewing and configuring heartbeat parameters

Cluster heartbeat parameters can be viewed and configured in iManager by selecting Cluster>Cluster Options>Cluster Object> Properties>Protocols, or from the Linux console by using the clustercommand (see Figure 5).

User space packages Service

dependencies novell-plugin-cluster-services-smp-1.8.0-20050324 novell-plugin-cluster-services-3.0.4-5 yast2-ncs-2.9.52-0.2 novell-nss adminfs heartbeat python-ldap

Figure 4. Some packages installed on a Linux server node during Novell Cluster Services installation

Figure 2. Novell iManager Cluster Options screen Copyright © 2005 Novell, Inc. All rights reserved. Used with permission.

Figure 3. Novell iManager Cluster Manager screen Copyright © 2005 Novell, Inc. All rights reserved. Used with permission.

(4)

The following heartbeat parameters can affect failover times: • QUORUMWAIT: Time in seconds that the cluster waits before

resources start to load

• HEARTBEAT: Time in seconds between transmits for all nodes in the cluster except the master node

• TOLERANCE: Time in seconds that the master node gives all other nodes in the cluster to signal that they are functioning • MASTERWATCHDOG: Time in seconds between transmits for the

master node in the cluster

• SLAVEWATCHDOG: Time in seconds that the slave nodes give the master node in the cluster to signal it is functioning • MAXRETRANSMITS: Maximum number of times transmits can

be attempted between the master node and slave nodes In the test configuration, the default parameters yielded failover times of approximately 45 seconds.

Managing storage resources

The CIFS protocol for storage resources does not work with Linux servers. To enable this functionality on Linux nodes in the test

configuration, the Dell test team manually configured a Samba server to share the cluster volume mount points. These shared mount points are configured differently by NetWare CIFS and Linux Samba; in the test configuration, the Samba node was set up to be equivalent to the existing NetWare CIFS configuration. After manual configuration, the disk volume was migrated using Novell Cluster Services. An example stanza in the /etc/samba/smb.conf file is shown in Figure 6.

To bring storage pool objects online, the storage resources pool must be active and the corresponding volume should be mounted and active on the master server node. These tasks can be per-formed using iManager or prompt commands: Entering volumesat a NetWare prompt ormount at a Linux prompt shows the active vol-umes on the node. For example, the Linux mount command returns the information shown in Figure 7.

The admin volume is a virtual volume used in creating and managing Novell Storage Services volumes. The admin volume is not persistent; each time the server boots, the volume is re-created.

Entering cat /proc/partitions in Linux shows the avail-able devices for the node. Novell provides the Enterprise Volume Management System plug-in evmsgui to allow device, partition, pool, and volume management. The evmsgui plug-in default loca-tion is /lib/evms/2.3.3/ncs-1.0.0.so.

Retiring old nodes

All services that are identical across platforms are expected to fail over from failing NetWare nodes to functioning Linux nodes in a cluster running Novell Cluster Services 1.8 or later. Services with identical configuration files and block data formats on both plat-forms are considered identical across platplat-forms. Apache 2.0.50 and MySQL 4.0.21 can migrate automatically. Services that cannot automatically fail over require manual migration between server platforms. Novell Cluster Services enables migration of the storage volume to the new platform, but configuration files for the service must be translated and created manually.

Successfully migrating from NetWare to Linux

Novell Cluster Services provides built-in tools and utilities to help make the migration from NetWare to Linux seamless. Even in the intermediate heterogeneous stage, the test cluster con-figuration maintained high availability and industry-acceptable failover times. Using the rolling migration procedure described in this article along with manual configuration steps for storage

oes-server02:~ # cluster info protocol

Cluster protocol internals Heartbeat: 1, Tolerance: 6

Master Watchdog: 1, Slave Watchdog: 8 Max Retransmits: 30

Figure 5. Heartbeat parameters shown using the Linux cluster command

[testvol7] comment = TESTVOL7 path = /media/nss/TESTVOL7/ writeable = yes browseable = yes guest ok = yes printable = no

Figure 6. Example stanza in the /etc/samba/smb.conf file

/dev/evms/TESTPOOL7 on /opt/novell/nss/mnt/.pools/TESTPOOL7 type nsspool (rw,name=TESTPOOL7) admin on /admin type nssadmin (rw)

TESTVOL7 on /media/nss/TESTVOL7 type nssvol (rw,name=TESTVOL7)

(5)

resources can help administrators reduce downtime when migrat-ing server clusters from Novell NetWare to Novell SUSE Linux Enterprise Server.

Cindy Stap has been in the computing field for 22 years and with Dell for

the past 9 years. She has holds Novell Master Certified NetWare Engineer (MCNE) and Microsoft Certified Systems Engineer (MCSE) certifications and has a bachelor’s degree in Computer Science from Almeda University.

Dean Oliver has been a senior analyst in the Dell Linux Operating System

Development Group for five years and holds Novell Certified Linux Engineer (CLE), Master Certified NetWare Engineer (MCNE), Red Hat Certified Engineer®

(RHCE®), and Microsoft Certified Systems Engineer (MCSE) certifications.

Ahmad Ali is a systems engineer on the Dell Linux Engineering team. He

has a B.S. in Electrical Engineering and an M.S. in Computer Engineering, both from Florida Atlantic University.

F O R M O R E I N F O R M AT I O N

Novell Cluster Services for NetWare 6.5 online documentation:

www.novell.com/documentation/ncs65/index.html

Novell Cluster Services for Linux online documentation:

www.novell.com/documentation/ncs_lx/index.html

Kosacek, Mike, and Dennis Dadey. “Implementing Novell Cluster Services.” Dell Power Solutions, Issue 4, 1999.

www.dell.com/content/topics/global.aspx/power/en/ ps4q99_novell

Lang, Richard. “Building a Scalable, Highly Available Novell Cluster Environment.” Dell Power Solutions, Issue 2, 2000.

www.dell.com/content/topics/global.aspx/power/en/ ps2q00_NetWare

Lang, Richard. “NetWare Cluster Services: Deployment Consider-ations and Tuning.” Dell Power Solutions, Issue 3, 2000. www.dell.com/content/topics/global.aspx/power/en/ ps3q00_lang

Pirich, Pete. “Upgrading a Novell NetWare Cluster to Dell SAN 3.0.” Dell Power Solutions, Issue 4, 2000. www.dell.com/ content/topics/global.aspx/power/en/ps4q00_pirich

References

Related documents

After converting all nodes in a former NetWare cluster to Linux, you might want to generate the cluster resource templates that are included with Novell Cluster Services for

Migration is the process of migrating services, file system data, and eDirectory information from an existing NetWare 6.5, Open Enterprise Server (OES) 1 Linux, OES 2 Linux or OES

It also describes how to implement these services in your network and configure the OpenWBEM Common Information Model Object Manager (CIMOM) on an Open Enterprise Server running

With Server Consolidation Utility 4.11, you can select a Windows domain as the source and copy Windows file system data along with users/groups, permissions, and ownership

If you decide you want to use only NTP as your time synchronization protocol, you can use the iManager > Time Synchronization > Migration plug-in to migrate NetWare 6.5 and

If you plan to use the Novell ® Storage Services TM (NSS) file system for data volumes on your OES Linux server, you might need to modify the Partitioning settings in the

This client allows all users of the workstation to install, delete, or administer printers and print jobs on the workstation, including printers and print jobs of other users. By

As the print services component of Novell ® Open Enterprise Server (OES), Novell iPrint provides a powerful and easy-to-implement printing solution that lets your network users