Large disk write queue
We have so called "high-traffic" websites and are investigating couchbase as backend for our cache. We created a Centos 5.8 server (Rackspace cloud server, 4GB mem, 160GB hdd) in the Rackspace cloud, installed Couchbase server 1.8.0r-55-g80f24f2 and created a Couchbase bucket with 3.19GB of storage.
The caching works fine for a few days, we noticed that the disk write queue goes up when our cache expires and that after a few minutes the disk write queue drops back to 0. But after a few days (approx. 3days) the diskwrite queue goes up to +/- 350k writes and stays above the 250k writes. This results in warnings in our monitoringsystem because the disk R/W load is >90%.
We would like to know if we can do something about this large disk write queue so there will be less load on the disks. Maybe somebody has experience with configuring Couchbase on the Rackspace and has some suggestions for us.
A screenshot of the graphs of our databucket can be found here: http://i.imgur.com/4UVa8.png