Skip to main content
Loading

Datacenter Replication Management

Versions 3.8 and above

As of Aerospike Server Enterprise Edition version 3.8, the Cross-Datacenter Replication (XDR) service is built in within the main Aerospike service (asd). As such it is always started and stopped when the Aerospike server is started and stopped.

Versions up to 3.7.5

The XDR daemon can be controlled using the SysV init script located at /etc/init.d/aerospike_xdr which supports the following options:

  • resume [default]
  • resume-nofailover
  • noresume
  • stop
  • status

Resume XDR from a Previous Execution [Default]

Starts XDR and will resume of shipping of the log from start of the log. Along with shipping data of current node it also takes responsibility of shipping of data for the nodes which are not present in the cluster currently. It will be able to ship data of a node which did not come up only if the digest log on this node has the shadow logs of it. This should be used when multiple nodes have gone down and all could not be brought back online. If the master and replica nodes corresponding to some data cannot be brought back online for any reason, obviously the data cannot be shipped because both the copies of the corresponding log will not be processed.

/etc/init.d/aerospike_xdr resume

Resume XDR but don't Ship for Others

Starts XDR with resumption of shipping of log from the point the last failure happened. However, the XDR will not take the responsibility of shipping on behalf of the nodes that did not come up. This should be used when a certain node(s) goes down and is brought up again and it is known that the shipping for the replica need not be done. If all the nodes are restarted each of them takes care of its own logs, all the log records should be covered. This is useful when doing a rolling upgrade or planned shutdown of the entire cluster when it is sure that all the nodes that are taken down can be brought up again.

XDR needs to be fault tolerant. So, when a machine is shutdown (or has a system failure), it should not lead to loss of data that is supposed to be shipped to the remote cluster. To handle this case, XDR writes the corresponding log records in the digest log (shadow log) at the replica site also along with the data. i.e, XDR records the information about what data has to be shipped to the remote cluster at the replica site also. But XDR does not ship using the shadow log unless the master node failed.

Once a node failed, its replica nodes will take over the responsibility of shipping the data that this node is earlier responsible for. This way XDR ensures that there is no data loss during a single node failure.

/etc/init.d/aerospike_xdr resume-nofailover

Start XDR with a Clear Backlog

Starts XDR afresh. Please note that once XDR starts with noresume all the old log records in the digest log are lost. This should be done when doing a fresh start. This can also be used in the event of a single node failure when that failed node is being restarted. When a single node fails, the responsibility of shipping the records belonging to it is taken over by XDR on a different node in the cluster. So, the XDR on the failed node, when it restarts, need not resume from the point at which it left. It can ignore them and continue to ship only the freshly written data.

/etc/init.d/aerospike_xdr noresume

Stop XDR

To shutdown the XDR service use the stop command:

/etc/init.d/aerospike_xdr stop

Get Running Status of XDR

To determine if the XDR service is currently running, use the status command:

/etc/init.d/aerospike_xdr status