- The system is live with this and one other site on Drupal 8 in a multi-site setup.
- The server is a LEMP server running Drupal 8 on Ubuntu 16.04.
- The OS is on a RAID 1 array, 1TB hdd (/dev/sda and /dev/sdb).
- The 256GB SSD is currently unformatted (/dev/sdc).
Disk Usage of Previous Web Server
root@hbw2live:/# du -m --one-file-system --max-depth=1|sort -g 0 ./dev 0 ./proc 0 ./run 0 ./sys 1 ./boot 1 ./lib64 1 ./lost+found 1 ./media 1 ./mnt 1 ./opt 1 ./selinux 1 ./srv 1 ./tmp 9 ./bin 9 ./sbin 10 ./etc 26 ./build 3517 ./usr 3856 ./lib 6832 ./home 8525 ./root 26294 ./var 49074 .
Desired Setup and Requirements
- Resilience to disk failure (hdd is already taken care of with RAID 1, but SSD is a single drive and I would not even consider using it were it not for the huge speed benefits).
- The root filesystem to stay on the hdd as I don't fancy taking a chance on moving it to another disk, nor do I want to risk the system on a single drive like that.
- Frequently-accessed directories in the course of serving a web page should be read from ssd for speed, but all writes must be replicated on the hdd for redundancy/failover/resilience.
- I was thinking of mounting selected directories on the ssd (etc, usr, lib, var) and syncing them to the hdd regularly - but point 3 above sounds almost like I should be using an automated file caching system that would use the ssd as a cache for frequently used files.