Microsoft Exchange 2000 Clustering Channing Heffney Support Professional Beta Team Microsoft Corpora

1
Microsoft Exchange 2000 ClusteringChanning
HeffneySupport ProfessionalBeta TeamMicrosoft
Corporation
2
Server Cluster Concept

• A cluster is a group of nodes working together as
  a single system to ensure that mission-critical
  applications and resources remain available to
  clients

A Cluster
3
Server Cluster Concept (2)

• Nodes are the physical servers that hold all the
  Exchange binary files
• Shared drive is an external disk drive physically
  plugged into all nodes in the cluster, which
  holds all the Exchange Data files (MDB and other
  files)

A Cluster
4
Server Cluster Concept (3)

• Network Resources
• Public network for client access to the nodes
• Internal network for inter-node communication
  (transmitting the nodes heartbeat) the network
  is optional for achieving higher availability

5
Server Cluster Concept (4)

• Cluster database the cluster database is
  located in the Windows 2000 registry on each
  cluster node. It contains information about all
  physical and logical elements in a cluster,
  including cluster objects, their properties, and
  configuration data.
• Quorum resource maintains the configuration
  data necessary for recovery of the cluster. This
  data, in the form of recovery logs, contains
  details of all of the changes that have been
  applied to the cluster database. The quorum is
  located on one of the external disk drive.

6
Server Cluster Concept (5)

• Each node's Cluster service maintains a
  consistent, updated image of the cluster database
  through the following mechanisms
• Global updates the Cluster service replicates
  changes to the cluster database uniformly across
  all nodes.
• Periodic checkpoint the Cluster service
  regularly checks each node's copy of the cluster
  database to ensure consistency.
• Quorum resource the Cluster service ensures
  that the quorum resource's recovery log contains
  the most recent cluster database information.

7
Server Cluster Concept (6)

• Forming a new cluster
• If a cluster does not exist, a node would try to
  gain control of the quorum resource to form a new
  cluster
• If successful, it uses the recovery logs in the
  quorum resource to update its cluster database
• Joining an existing cluster
• If a cluster exists, a node tries to communicate
  with another cluster node to join a cluster
• Then it would update its copy of the cluster
  database from that on the other active node

8
Server Cluster Concept (7)

• Resources any Windows or Application resources
  or service/instance
• Group a group is a collection of cluster
  resources which define the units of failover
• Failover moving a group of resources from one
  node to another
• FailBack moving a group of resources back,
  after a failover, to the original node when it
  becomes available

9
Virtual Server

• A Virtual Server (Exchange) is a group that
  contains
• An IP address resource
• A network name resource
• Shared disk resource
• Exchange System Attendant

10
Exchange Virtual Server

• An Exchange Virtual Server (EVS) would be a
  Virtual Server containing Exchange resources and
  a shared external disk drive containing the data
  and log files
• Each node has a copy of the Exchange binaries
• Each Exchange resource in a group represents an
  instance of an Exchange service, but the service
  itself, for example, an IMAP4 resource, is
  actually an IMAP4 virtual server instance

11
Exchange Virtual Server (2)
12
Exchange Virtual Server (3)
EVS Data
Local Drive Exchange Binaries
13
Active/Passive Cluster
Node One
Node Two
Quorum
EVS1
IDLE
EVS2
EVS3
14
Active/Active Cluster
Node One
Node Two
Quorum
EVS1
EVS2
EVS3
15
Active/Active Cluster (2)

• Active/Active and Active/Passive are just terms
  to describe the stage of Exchange cluster
  capability

16
Exchange Virtual Server
EVS1 Data
EVS2 Data
Local Drive Exchange Binaries
Local Drive Exchange Binaries
17
IsAlive

• IsAlive is a mechanism to determine whether a
  service is alive or dead. If a service is alive,
  no operation takes place if its dead, the
  cluster service will either restart the dead
  service or fail over the Virtual Server. The
  cluster resource monitor will periodically call
  the IsAlive function, to determine whether a
  service is alive or dead.

18
IsAlive Reference

• Store Exres will use the same account that Mad
  uses to do a MAPI logon and open the Inbox. Its
  alive if a logon and open Inbox are successful.
• IMAP4 Exres will make a IMAP4 connection to the
  IMAP4 server, if a welcome banner is returned,
  the IMAP4 server is alive. Exres will read the
  metabase for the port number to use for this
  connection. Each IMAP4 instance might have a
  default port number or different one.
  Furthermore, each Virtual Server can have
  multiple IMAP4 instances, therefore each of these
  instances will have its unique port number.

19
IsAlive Reference (2)

• POP3 Exres will make a POP3 connection to the
  POP3 server, if a welcome banner is returned, the
  POP3 server is alive. Please see the IMAP4
  description, its very similar to IMAP4.
• SMTP Exres will make an SMTP connection to the
  SMTP server, if a welcome banner is returned, the
  SMTP server is alive. Please see the IMAP4
  description, its very similar to IMAP4.
• DAV Exres will connect to the DAV server and
  send Track verb to the server. If the response
  is a success, the server is alive.

20
IsAlive Reference (3)

• Routing Exres will make a RPC call into the
  routing service, the routing service will make
  its own determination whether they are alive or
  dead and return appropriately.
• Mad Exres will make an RPC call into the Mad
  service, the Mad service will make its own
  determination whether they are alive or dead and
  return appropriately.
• SRS Is not supported on a cluster.
• Connectors Exres will check with the Windows
  2000 Service Manager to determine if the
  Connector service is up or down.

21
Registry Replication

• Checkpoint A pointer to all registry keys
  associated with a resource in the cluster the
  checkpoint sits on the quorum
• When a failover of a group occurs, the quorum
  would replicate the registry key from one node to
  another according to the checkpoints

22
Failover Behavior

• In Exchange 2000, even if you manually fail over
  an Exchange Virtual Server 1 (EVS1) from Node A
  to Node B, the Exchange service in Node A will
  remain running. Only instances for EVS1 will
  stop.
• MTA and Connectors services, which only support
  one instance per cluster, would stop when the EVS
  that owns them fails over.
• SMTP service would stop every time one of its
  instances goes offline. The remaining instances
  on the node would restart the SMTP service again.

23
N-Node Cluster

• All concepts in this presentation apply to
  N-node clusters

24
Components Not Supported on a Cluster

• Conferencing
• Instant Messaging
• KMS
• Foreign mail connectors (Notes, GroupWise, MS
  Mail, ccMail)
• SRS
• NNTP

25
Components That Are Not Active/Active

• Chat
• Exchange 2000 MTA

26
Active/Active

• SMTP, POP, IMAP, HTTP
• Store
• System Attendant
• Context Indexing

27
Things to Know!

• Bug 160312
• Bug 157414
• There can be only one public folder store in the
  cluster.
• Remember! The limitation of storage groups does
  apply to clusters as well.
• No virtual server can support more than four
  storage groups at one time (failover).

28
Setup Instructions

• Setup instructions can be found in
• Knowledge Base article Q263272, XADM How to Set
  Up Exchange 2000 Server on a Windows Cluster
• In the C20_Clustering.rtf file on the Exchange
  2000 CD

29
(No Transcript)