Does anyone know how to make cbworkloadgen use random reads?
It seems to always be reading the same documents as I got my cluster down to 0% resident ratio and I still cannot get cache misses…
I’m not sure that’s a feature of cbworkloadgen, but you’ll find the python source on github and it should be pretty straightforward to modify.
Also, note that libcouchbase includes a tool called pillowfight as part of the cbc tools which does do random reads so you can generate cache misses. There’s a slightly old writeup on the blog.
Note that pillowfight should be faster as cbworkloadgen is pure python to facilitate building .exe files for Windows to not need another runtime.
Maybe @mnunberg can write up a fresher blog on pillowfight and friends since it’s been enhanced a bit since 2012.