SSD Optimizations

[joaquincasares]

Write ssd logic and include these commands when ssds are detected:

echo noop > /sys/block/sda/queue/scheduler
          
echo 0 > /sys/block/sda/queue/read_ahead_kb
          
echo 0 > /sys/block/sda/queue/rotational

[ltagliamonte]

@joaquincasares what do you think to also add a weekly scheduled fstrim?

[mlococo]

@joaquincasares is no longer active on the project. I'm doing most maintenance for now.

I'm not current on the latest advice for SSD's, and I believe trim behavior has changed in recent years. I'd want any of these optimizations to be backed by a citation from a credible source that includes some discussion of tradeoffs.

I'm also somewhat wary of overoptimization. While ComboAMI is a great and easy way to kick the tires on a cluster, it's not well suited for running a prod cluster. It provides no assistance during upgrades, datacenter expansions, repairs, or any of a million other maintenance tasks. If you're handling those other aspects of your cluster's maintenance, firing off trims isn't going to be a big incremental effort. If you're not handling those other aspects of your cluster, ComboAMI doing trims for you isn't going to keep your cluster healthy for long.