Currently if a remote replica is busy the replication protocol will go into a “back off” state, and it will retry to send it updates at the next interval of the backoff timer. By default the timer starts at 3 seconds, and has a maximum wait period of 5 minutes. These default settings maybe not be sufficient under certain circumstances. In 389-ds-base-1.3.1 you can now configure the minimum and maximum wait times.
In environments where there is heavy replication traffic, and updates need to be sent as fast as possible, having a maximum retry time of 5 minutes is too long.
Two new configuration attributes have been added to the replication configuration entry(per backend):
cn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config
...
...
nsds5ReplicaBackoffMin <value in seconds> default is 3 seconds
nsds5ReplicaBackoffMax <value in seconds> default is 300 seconds
No additional requirements are needed.
The configuration must be handled through CLI tools.
The configuration settings can be applied while the server is online, and do not require a server restart. If invalid settings are used, then the default values will be used instead.
Impacts the back-off timers in the incremental update replication protocol.
No impact on upgrades/updates.
Feature was added to 389-ds-base-1.3.1.
No external impact.
Mark Reynolds mreynolds@redhat.com