Very strange Aerospike behaviour - completely unknown cause


#21

Hmm, that’s valid - sections of our network are running on 10GbE but others are only 1 - I’ll get DevOps to look into it and see if this could be a possibility…


#22

@kporter - I have run those commands and we’ve looked at a few other things on the network side of things and it’s a possibility that we’re hitting a 1GbE limit on some services. Are there any commands I can run on Aerospike server via benchmarking or microbenchmarking to see how much data is being sent from the server, either as a value per second (such as Kb/sec) or as a total so we can see what we’re sending to the clients.


#23

No, we do not have any client throughput metrics. There are intra-cluster throughput metrics that can be seen in the logs or from the info protocol. Specifically fabric_rw_send_rate and fabric_rw_recv_rate show the rate for transaction handling (such as replica writes or duplicate resolution).