Various issues with a geo-redundant Cassandra setup with multiple data centers

Affected versions

Geo-redundant or site-redundant Cassandra Cluster setups using multiple data centers

Cause

DataMiner maintains references to multiple contact points in the DB.xml configuration file, which may be spread across two or more data centers in Cassandra. DataMiner automatically detects the data center of the first contact point in DB.xml and evenly distributes queries across all nodes within that data center. If a query consistently fails, the policy will not attempt to connect with nodes in other data centers. In such cases, a DataMiner offload alarm will be triggered, and DataMiner will initiate the process of offloading.

Issues

Below you can find an overview of issues related to this Cassandra Cluster setup:

Issue	Affected versions	Workaround	Fix
DataMiner does not automatically reconnect to a different Cassandra data center when queries fail consistently.	All Cassandra Cluster setups	N/A	Install DataMiner 10.3.0 [CU8]/10.3.11
DataMiner goes into offload mode after losing only one node. See: Cassandra disconnects after loss of a single node	All Cassandra Cluster setups	N/A	Install DataMiner 10.3.0 [CU7]/10.3.10.
When the connection with the Cassandra cluster is temporarily lost, data is not offloaded to database offload files and restored to the database afterwards, causing data loss for the duration of the connection issue. See: Cassandra cluster data not offloaded while database is unavailable	All Cassandra Cluster setups	N/A	Install DataMiner 10.2.0 [CU18]/10.3.0 [CU6]/10.3.9.