Dear DOJO,
This is an interesting question related to a DataMiner cluster of 7 DMAs running version 10.1. When full DataMiner backups are executed, one would expect the Cassandra DB disk size to be similar to the size of the DataMiner Full Backup file. Example below is for a DMA with this expected behavior (DMA1):
We are seeing a different behavior in two DMAs of this same cluster. See screenshot below for DMA 5:
As you can see, this DMA 5 backup was completed with no errors with 50% of the size of the Cassandra DB. There are other DMAs for which the backup is 25%~30% of the Cassandra DB.
How can this size difference be explained?
As a follow-up question, we'd also like to know if there is a way to query a Cassandra DB to find the list of elements that takes more space in the DB. We know the amount of KPIs stored for historical alarm and trending purposes are really the heavy contributors of this DB disk size.
Thanks!
Hi Luiz,
The Cassandra folder contains more than only the ssTables. Most likely you have some snapshots that are still present in your Cassandra folder. Can you have a look inside one of the table folders if there are any snapshots available (e.g. E:\ProgramData\Cassandra\SLDMADB\data-[tableID]\snapshots)? To get rid of your snapshots you can perform a nodetool clearsnapshot, however there is a know bug on Windows OS that the Cassandra process does not always releases the lock correctly. Then a Cassandra restart is needed after the nodetool clearsnapshot to get rid of the snapshots.