Close to breaching memory HWM when using XDR

Since enabling XDR memory usage has gone up and we are nearing a possible breach of HWM. my XDR configuration is largely the default and i was wondering if there were parameters I should consider tweaking, e.g., would turning on compression help?

I would suggest opening a case with Aerospike Enterprise Support as this would require gathering of logs to check details. Compression for XDR only affects what would be put on the wire and should not impact memory much. Assuming running version 5.X, the XDR transaction queues (transaction-queue-limit) would use some memory (depending on the cluster size as it would be on a per partition level).

Thanks, I did. The conclusion was that it was the index that was actually taking up space. On closer examination the source site was actually larger than the destination. Now I am trying to figure out why when I turn on XDR I am seeing a significant decrease in client_connections and file_handles… Of course step one is to start the experiments all over again with a resized destination…

Thanks for providing an update here. Good to know.

This topic was automatically closed 84 days after the last reply. New replies are no longer allowed.