After upgrading to CE 3.0.1 we are seeing memcached use twice as much CPU. We have a ten node cluster and all nodes are showing the same behavior. Are others seeing this behavior? Is it expected?
Can you tell us a bit more about your configuration?
OS and Version?
How many buckets are on the cluster?
How many views and design documents are defined?
Are you using XDCR?
CentOS 6.6, kernel 2.6.32-504.3.3.el6.x86_64
ten nodes, 4GB per node
one bucket (default)
no views or design documents
59896608 items, 1 replica
during business hours we average around 50K ops/second
Also it wasn’t an in place upgrade, we started the cluster from scratch
The CPU usage stays consistent even during off hours which seems strange. What does that memcached process do that would cause it to consume cycles when there are very few requests?
There are a few things that will cause CPU utilization even when the system is idle: for example: heart-beats: messages among nodes or stats: statistics collection. However these don’t cost you more when you have a busier system.
I took your first post as you are seeing the CPU utilization double under load: is that right? or are you concerned more about the idle system situation?