Details
-
Bug
-
Resolution: Done
-
P2: Important
-
master, production
-
None
Description
Currently we use default Ubuntu setup for our hosts, while they have quite unusual IO characteristics. We could consider increasing the dirty page counts and disk cache in general. If it is big enough we could also increase disk life time, as some VMs would never touch the local storage.
Two things needs to be done here:
1. monitor /proc/vmstat for all hosts
2. change the default parameters