We are currently running CE v3.16.0.6 and use GitHub - aerospike-community/aerospike-graphite: Aerospike monitoring for Graphite - a community driven open source project to send aerospike metrics to our graphite setup; these metrics ultimately get visualized in dashboards in Grafana. I have dashboard that is akin to the Cluster Throughput panel on AMC (e.g. it breaks down the reads and writes per second for the cluster.). I was using the latency.*.read.ops_per_sec and latency.*.write.ops_per_sec for this dashboard and when we had an application that changed the client to performing batch reads. I noticed a drop in latency.*.read.ops_per_sec. Is there a reason why batch reads are not included the latency.*.read.ops_per_sec? I see them being reported separately under the batch_sub_read_(success|error|timeout) metrics. Should I just update my dashboard to add the batch_sub_read_(success|error|timeout) metrics to the latency.*.read.ops_per_sec or track those separately?
The batch reads are indeed separate from the individual read transactions when it comes to metrics. I am not sure whether there is a reason other than the transaction flow being different at the high level. It can also be an advantage to have this separation which can help troubleshoot potential bottlenecks.
This topic was automatically closed 84 days after the last reply. New replies are no longer allowed.