I typically see one or two boxes in my clusters shouldering the majority of the work. Here’s one box doing way more than the other three in a 4-node cluster:
http://aerospike.con.com/asd.png
The load averages on the individual boxes follow this:
20:11:58 up 1 day, 21:43, 1 user, load average: 3.03, 3.08, 3.11
20:11:58 up 1 day, 21:42, 1 user, load average: 0.31, 0.33, 0.39
20:11:58 up 204 days, 20:43, 1 user, load average: 0.36, 0.39, 0.66
20:11:58 up 36 days, 23:36, 1 user, load average: 0.77, 0.68, 0.77
Why is there such skewed loading between nodes? Is there something I can do to analyze this better?
This is with the 3.6.0-1 Community server running on late-model Amazon AMI’s (el6).