The Cluster Consistency Monitor is intended to be an adaptive
toolkit. This means that during the initial run, all the generic
resources like the shared volume groups and the file systems of
an application, which may influence the proper operation of that
application in the cluster, are specified in the cluster resource
DB. Then, individual configurations for a certain application or
in a certain customer environment can be added whenever necessary
by simply adding the resource definition to the resource DB. This
flexibility enables the Monitor to work with all kinds of applications.
Even in uniform applications like SAP R/3, many customers have different
individual setups and need flexibility to address this.
The Cluster Consistency Monitor must be run periodically in
order to maintain the cluster failover capability. The current implementation
of the monitor supports a "Learning Mode" which gives detailed Information
about a discovered failure as well as some hints how to fix the
problem. The knowledge of configuring and maintaining a cluster
has been put in the Monitor. On the other side, in case a cluster
is well tuned, this feature may be switched off and the monitor
checks only on changes. Of course, potential problems may be fed
into a management system like HP ITO.
In a small installation consisting only on a few clusters,
the output should be delivered in an ASCII report. It might be however
useful to have a central maintenance server in a large IT infrastructure.
The current implementation of the monitor supports an ASCII report
for standalone operations as well as HTML reports for centralized infrastructure
repositories.
The basic function of the Cluster Consistency Monitor is to
compare resource configurations of nodes. Each node that is monitored
runs a program which reads a configuration profile about named resources
and creates a resource database. These resource databases are then compared
by another program and possible error conditions are presented in
an ASCII or HTML report which can be used for troubleshooting.
Why Cluster Consistency Monitor is Needed |
 |
Clusters exist to maintain the operation of an application
even in case a hardware resource failure. The SAP R/3 environment
is very complex, depending on a number of resources on multiple
servers. Some of these resources run on the nodes in the cluster,
while others are on non-cluster nodes. Figure 1-11 “Cluster Configuration for an
SAP R/3 Application” depicts
a typical view of a cluster configuration for an SAP R/3 application.
Because of this complexity, the precise definition of resources
on each server in and out of the cluster must be managed carefully.
The Cluster Consistency Monitor is a tool
that assists in maintaining consistent relationships among all component
servers.
Challenges in a Cluster Environment
While we typically find the Central Instance and the Database
of a SAP R/3 System along with some important filesystems shared
by NFS, there are a number of other dialogue instances associated
with the cluster. Figure 1-12 “Dialogue Instances of the Cluster
Configuration” shows the applications
resources, filesystems, volume groups, IP name and process control
have been put in a package and outside of a cluster configuration
for a SAP R/3 application.The start and stop of these instances
may also be controlled by the SGeSAP solution.
The package controls these resources and relocates it upon
a node failure to another node. While important application components
are in the shared devices, the definition for the shared devices,
as well as the configuration of the relocatable IP address have
to be maintained manually on each node capable of running this application.
Features of the Cluster Consistency Monitor for SGeSAP
Environments |
|
The implemented features of the Cluster Consistency Monitor
are as follows:
The information to examine is configurable
in order to support all kind of applications.
Due to the flexibility in configuring resources,
any kind of cluster may be monitored. This includes the basic MC/ServiceGuard
Cluster, MetroCluster, and installations of shadow DB and HP Sommersault units
as well.
Updates on the configurations are easy.
The compare run may be performed automatically as
a background task or individually from the command line.
The Cluster Consistency Monitor may be used to trigger
an event for a management system like ITO upon an error.
Results are provided in a report. Output format
of this report may be HTML or ASCII.
Output reports may contain error interpretation
text in order to help during interpretation of the compare.
All resources configured in a package may be easily
added to the resource DB by using the statement ANALYSE PACKAGE upon creation of the Resource DB.
It is possible to suppress acknowledged error messages.
Changes in a cluster setup, which may be OK, and may be acknowledged
by implementing an entry in an error exclusion table in order to suppress
them. No trigger to external systems would be created on suppressed
messages.
The Cluster Consistency Monitor detects and reports
all kinds of setup problems of the monitor itself.
The operation of the cluster monitor does not impact
applications, which are currently running due to the resource consumption
of the monitor is very low. Typically, the comparator takes less
than 10 seconds in comparing two nodes, which includes the creation
of the resource DB.
The use of the Cluster Consistency Monitor avoids
fail over tests. Due to the fact, that on a fail over test, only
the startup of the package on a secondary node is examined, potential
problems which may arise in case the application runs for a longer
period on that node are not recognized. The Cluster Consistency
Monitor detects even that class of problems as it simulates the
view an application would have toward the configurations of operating
systems.
The storage format of the resource DB is very dense.
It allows fast transfer even over wide area networks. Due to that
density, memory consumption during the comparator run or storage
consumption for the resource DB is very low.
The current implementation supports HP-UX 10.20
and 11.00, 32 Bit and 64 Bit.
Installation and Setup of the Cluster Consistency Monitor |
|
Refer to OS998 in Chapter 2 for information on installing
the Cluster Consistency Monitor. More detail about installation
and setup of the monitor is found in the monitor handbook located
after installation in the following directory: /opt/cctool/doc/CCmon_Handbook.doc.
Printable version
