Hard Drive Strategy SO CONFUSED

[Amorbellum]

im constantly running into this point where i just, literally have no idea how to manage my hard drives and everyone has a different opinion

 

i have a standard array. im running plex and one vm so far, two eventually

 

i have a 500gb ssd, a 1tb ssd.

 

my plan was to use the 500gb as the cache disk. when copying from my old server to my shares through the cache, it would fill up and copying would stop. SUPER frusterating. so i turned off cache altogether. i thought when it was full it would pass right to the array?
 

and should app data, and vms be on the cache? or on a separate disk? i wanted to use the 1tb ssd as my landing disk for sonarr/deluge. so should i be getting another ssd for vm? the vm im hoping to pass through a video card at some point for light gaming

 

so what is everyone doing. is a 500gb cache good enough? then why do i have to run the mover, shouldnt files jsut go past it when its full?

do vms needs their own disk?

does nything else? plex? appdata?

i cant find straight answers on this

[JonathanM]

6 minutes ago, Amorbellum said:

i thought when it was full it would pass right to the array?

Correct, as long as unraid knows it's full. You have to tell unraid how much free space to hold back to start passing new writes to the parity protected portion. Settings, global share settings, cache settings, min free space.

 

As far as cache disk usage, it's different for different usage, so you will get a variety of opinions. If you are constantly filling the cache, you probably need to change your strategy.

[Amorbellum]

and your vms? and your downloads?

 

would it make more sense to use an unassigned conventional disk, like just some old western digital, as my download landing? that frees up my 1tb to use for cache

 

does it make sense to keep everything else on cache then?

Edited May 30, 2019 by Amorbellum

[tjb_altf4]

4 hours ago, Amorbellum said:

and your vms? and your downloads?

 

would it make more sense to use an unassigned conventional disk, like just some old western digital, as my download landing? that frees up my 1tb to use for cache

 

does it make sense to keep everything else on cache then?

My system is set this way and I've had good performance from this configuration.

 

My downloads (torrents) land on an unassigned HDD, it then gets cleaned up etc and sent to the array (via cache and mover).

I find this takes a lot of disk IO traffic away from array and cache so docker and vm performance remains consistent.

note: my VMs and docker live on the cache.

[testdasi]

14 hours ago, Amorbellum said:

and your vms? and your downloads?

 

would it make more sense to use an unassigned conventional disk, like just some old western digital, as my download landing? that frees up my 1tb to use for cache

 

does it make sense to keep everything else on cache then?

Unraid cache has evolved from being just a glorified temp space to being an integral component (e.g. where to save VM vdisks, docker, appdata, libvirt etc.) so I would recommend to NOT use it as temp space. Using it for temp space will wear it down very quickly and when (not if) the cache drive fails, you will have a lot of "fun" getting things back up running.

The most common mistake I see newcomers to unRAID make is to automatically use Cache = Yes (i.e. the original "glorified temp space" intention) without considering any other alternative and if fast write speed is even critical to begin with. For example, I am totally content with 75MB/s write speed backing things up from my laptop to the Unraid server because it's done at night and/or in the background when I'm doing other things. When needing to urgently back up a 100GB project, I turn on reconstruct write (aka turbo write) and be happy with it because gigabit network maxes out at 125MB/s, bottlenecking the max 150MB/s write speed with turbo write.

 

Basically, before using cache for any share, ask yourself - do I really need 300MB/s write speed? If you ask yourself that question, you should naturally come to the conclusion that download landing shouldn't be on cache. Unassigned device is your best bet but don't use anything expensive.

 

On a side note, people talk about wearing out SSD but remember, repeated write will wear HDD out too, even more so considering the mechanical nature of it. Wearing out an SSD usually just means increasing number of bad sectors while the rest of the drive is still very much usable (unless you have Intel SSD). A worn-out HDD will usually catastrophically fail.

[Glassed Silver]

13 hours ago, testdasi said:

Unraid cache has evolved from being just a glorified temp space to being an integral component (e.g. where to save VM vdisks, docker, appdata, libvirt etc.) so I would recommend to NOT use it as temp space. Using it for temp space will wear it down very quickly and when (not if) the cache drive fails, you will have a lot of "fun" getting things back up running.

The most common mistake I see newcomers to unRAID make is to automatically use Cache = Yes (i.e. the original "glorified temp space" intention) without considering any other alternative and if fast write speed is even critical to begin with. For example, I am totally content with 75MB/s write speed backing things up from my laptop to the Unraid server because it's done at night and/or in the background when I'm doing other things. When needing to urgently back up a 100GB project, I turn on reconstruct write (aka turbo write) and be happy with it because gigabit network maxes out at 125MB/s, bottlenecking the max 150MB/s write speed with turbo write.

 

Basically, before using cache for any share, ask yourself - do I really need 300MB/s write speed? If you ask yourself that question, you should naturally come to the conclusion that download landing shouldn't be on cache. Unassigned device is your best bet but don't use anything expensive.

 

On a side note, people talk about wearing out SSD but remember, repeated write will wear HDD out too, even more so considering the mechanical nature of it. Wearing out an SSD usually just means increasing number of bad sectors while the rest of the drive is still very much usable (unless you have Intel SSD). A worn-out HDD will usually catastrophically fail.

Maybe it's about time unRAID reigns in a new concept. In my humble opinion a cache should be a cache and the fast permanent storage should be an entirely different drive. Ideally with an automated way to protect it with parity as well, maybe not on the fly, but at least to save snapshots to the array.

 

All drives fail eventually and do I really want to go through the hassle to re-deploy my VM and docker storage? No, not really...