How to gather transaction queue status?


#1

Hi

I am trying to analyse the performance bottleneck for my cluster (8 node , disk + DRAM mode) As soon as the the read+write load on individual node goes beyond a particular threshold( 5500 tps), performace starts degrading drastically. And this has started to happen since I moved to new I3 series boxes of AWS from I2 series on which it was running fine

I am trying to gather all metrics like cpu , disk io ,etc. in relation to tps and see what is causing the slowness. I would like to know the length of my transaction queues also

I will also try and increase the thread per queue as my cpu utilisation is less than 20% while max disk utilization is 60%,

using 3.9.1 version


#2

CPU and disk io collection would have to be done from OS level tools and not AS I believe. For transaction queue, this is logged frequently and also available under stats http://www.aerospike.com/docs/reference/metrics#tsvc_queue http://www.aerospike.com/docs/reference/serverlogmessages defrag -q write-q tsvc-q