- 论坛徽章:
- 0
|
When you configure an HACMP cluster, configuration data is stored in
HACMP-specific object classes in the ODM. The HACMP for AIX ODM object
classes are stored in the default configuration directory (DCD), /etc/es/objrepos.
It is possible to perform certain changes to both the cluster topology and cluster
resources while the cluster is running. This operation is called Dynamic
Automatic Reconfiguration Event (DARE).
Prior to HACMP V5.1, making changes to the cluster topology and cluster
resources could be very time consuming, since this required running multiple
DARE operations in the cluster.
It was not possible to perform dynamic reconfiguration changes to both
resources and topology during the same operation.
HACMP V5.1 allows a combination of resource and topology changes via one
dynamic reconfiguration operation.
DARE operation
The dynamic reconfiguration requires changes to be made to HACMP ODM
classes on all nodes in the cluster. In previous HACMP versions, whenever a
dynamic configuration change was performed, multiple remote connections to
the other nodes in the cluster were issued for each class that had to be modified.
In the current version, due to the cluster communication infrastructure
(clcomdES), the connection to remote nodes is already opened and maintained
as active, so remote operations are much faster.
At cluster start-up, HACMP copies its ODM classes into a separate directory
called the Active Configuration Directory (ACD). While a cluster is running, the
HACMP daemons, scripts, and utilities reference the ODM data stored in the
active configuration directory (ACD) in the ODM.
If you synchronize the cluster topology or cluster resources definition while the
Cluster Manager is running on the local node, this action triggers a dynamic
reconfiguration (DARE) event.
Restriction: If you have sites defined in the cluster, you cannot perform
changes to the cluster resources or topology using DARE. This is due to the
fact that during reconfiguration, the secondary (remote) site may falsely detect
a primary site failure and initiate the take-over process.
In a dynamic reconfiguration event, the ODM data in the Default Configuration
Directories (DCDs) on all cluster nodes is gathered on the node that initiates the
reconfiguration process in a Staging Configuration Directory (SCD). The HACMP
ODM classes from remote nodes are gathered on the local node and a
checksum with a time stamp is performed and sent back to the nodes.
The configuration changes are performed, the ODM classes are updated in the
SCD, and then are sent back to the originating nodes. The checksum is verified
to avoid data corruption if the verification/synchronization has been started from
more than one node and, if the checksum is correct, the ODM classes in the
DCD on the target nodes are updated (overwritten).
In the final step, the ODM data in the ACD is overwritten with the new
configuration data. The HACMP daemons are refreshed so that the new
configuration becomes the currently active configuration
本文来自ChinaUnix博客,如果查看原文请点:http://blog.chinaunix.net/u/9663/showart_49684.html |
|