Starting ASD on a node with Junk data


#1

Summary:

If ASD is started on a new user data node that hasn’t had its ssds wiped, how does ASD react to the junk data that is in there?

Resolution:

Ideally, we recommend we have the SSD’s wiped clean when adding a new node to a cluster to avoid any conflict with existing data. If the previous data wasn’t from an Aerospike instance than it is likly safe to use without issue. If it were from an Aerospike cluster then there are a couple scenarios to consider.

Two scenarios:

Scenario 1: Existing data is from another Aerospike Cluster with the same Namespace Name:

In this scenario, adding the node to the cluster will result in a portion of the data from the old cluster being replicated to this cluster.

Scenario 2: Existing data is from another Aerospike Cluster with the different Namespace Name:

In this scenario, Aerospike should recognize that the disk is from another namespace and will terminate server start.

Workaround:

There are situations where it isn’t desired to load data from disk (such as testing). For these cases we provide the cold-start-empty configuration parameter. This instructs Aerospike to not read the disk during a coldstart which. Once this is set for a namespace it isn’t recommended to remove the parameter.