Replication Protocol Timeout


Overview

When stopping the server, disabling replication, or removing a replication agreement, there is a timeout on how long to wait before stopping replication when the server is under load. Prior to 1.3.1, the server could wait up to 20 minutes which would look like a hang. In 1.3.1 you can now configure this timeout, and the default timeout has been reduced to 2 minutes.

Use Cases

There could be cases where a timeout of 2 minutes is too long, or not long enough. There could also be situations where a particular agreement should be given more time before ending a replication session during a shutdown.

Design

This attribute nsds5ReplicaProtocolTimeout can be added to the main replica configuration entry: e.g. Zcn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config

Or, to the a replication agreement. The replication agreement protocol timeout will override the timeout set in main replica configuration entry. This allows some different timeout for different replication agreements.

Implementation

None.

Feature Management

CLI

Major configuration options

nsds5ReplicaProtocolTimeout: <value in seconds> The default is 120 seconds.

This can be added the main replication config entry for a backend: cn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config

Or to any replication agreement: cn=agmt to remote replica,cn=replica,cn=dc\3Dexample\2Cdc\3Dcom,cn=mapping tree,cn=config

Replication

If a replication session is in progress the new timeout will abort that session and allow the server to shutdown.

Updates and Upgrades

No impact.

Dependencies

No dependencies.

External Impact

No external impact.

RFE Author

Mark Reynolds mreynolds@redhat.com

Last modified on 7 August 2014