The following counters are relevant to database replication:
The average number of bytes needed to replicate an object from this machine. Note that this excludes BLOB data.
The average number of bytes transmitted to a machine during each replication (a replication typically consists of many objects). This excludes BLOB data. This counter can be used to help determine network impact.
The number of 'failed' notifications to connected machines. Note: It is quite normal for notifications to fail, if for example the connected machine is not available when the notification attempt is made. However a steadily increasing number of failed notifications may indicate a communication problem related to the local machine.
The number of objects waiting to be replicated.
The number of events replicated to this machine since the infrastructure service was started.
The total number of events replicated to this machine per second.
Total number of events replicated to the parent machine since the infrastructure service was started.
The total number of events replicated to the parent machine per second.
This is the current (active) number of connections that are applying replication data to this machine. This counter can be used in conjunction with the 'Incoming Connections Pending' to help tune the policy settings that govern incoming connection handling.
This counter shows the number of connected machines that are waiting to replicate data to this machine. A value of 0 indicates that this server has no replication backlog, whereas a value consistently near the number of connected machines may indicate that the machine is overworked.
The size (in bytes) of the last batch of replicated objects sent from this machine. For application data, if this value is consistently close to the policy specified batch size then it indicates that this machine is processing data faster than its parent machine. A consistently small value may indicate that this machine cannot pass data to the parent quickly enough.
The size (in cache objects) of the last batch of replicated objects sent from this machine.
The last time (in milliseconds) needed to notify all connected machines that they have data to replicate. This can be compared against the average time (Notification Time) to help verify notification related problems and performance.
The average time (in milliseconds) needed to notify all connected machines that they have data to replicate. This counter can be used to determine how much time a machine actually spends 'notifying' connected machines, and can be used to tune the policy notification intervals to acceptable values.
The number of objects replicated into the local infrastructure.
The rate at which objects are being replicated into the local infrastructure.
The number of objects replicated out of the local infrastructure.
The rate at which objects are being replicated out of the local infrastructure.
This is the count of active machines currently reading replication data from this machine. It can be used to help identify periods of heavy load on CMS and Gateway machines caused by large amounts of replication traffic.
When the replication notification scheduler becomes active, it must determine which machines require replication notifications to be generated. This counter is the number of connected machines that still need to be processed by the notification scheduler in the current notification period. It can be used to show the behavior of the notification scheduler over a notification period. A value of 0 means that the scheduler is currently inactive, waiting for the next scheduled notification.
The average time (in milliseconds) taken by all connected machines to read replication data from this machine.
The total number of bytes replicated to this machine in the last second. This total relates to database objects only, and excludes any BLOB data.
The total number of bytes replicated to the parent machine in the last second. This total relates to database objects only, and excludes any BLOB data.
The number of successful notifications made to connected machines. This value should increase (by 1 for each notified machine) when the notification scheduler is active.
Infrastructure data is replicated by notifying 'key' information of all object changes to all connected machines. However client and utility machines only request full details of the changes that apply to themselves. This counter (maintained by the publisher – usually the parent machine) indicates (as a percentage) how many of the replication objects are actually requested by the connected machines.
|
Copyright © 2014 CA.
All rights reserved.
|
|