I've had my Dockers on the cache since I first started using Dockers. The main use is for the Plex Server and keeping the Plex Metedata files, Indices etc on cache certainly speeds things up.

 

However, I have a problem. My Plex installation probably now uses half of the 256GB SSD cache drive and on more than one occasion I have managed to fill the balance of the cache with Media files in transit. By which I mean Mover hasn't got around to cleaning up the cache. Whilst this isn't a problem for those files, I'm pretty certain it does cause problems for the Dockers which may face a situation where the cache is full. I did see there was a script to invoke the mover and that this could run using the CA User Scripts. However, this seems to be a very clumsy solution to a problem that is inherent in using the cache for Dockers.

 

So I was wondering whether anyone else has encountered this issue, whether they have adopted the script solution or is there another more elegant way of resolving this. I guess one way would be to install another SSD which is only used for the Docker files in which case it would be easy to monitor use of the Docker storage. Or perhaps what would be a solution would be for 'cache only' files to be allowed to overflow onto the unRaid drives should the cache become full. In which case the Mover would then cleanup 'cache only' files when space became available on the cache?

 

Any thoughts?

 

 

If pure write speed of the media files isn't your top priority, set the media shares to not use the cache drive.  (After an hour or two of downloading, does it really matter if it takes the copy an extra minute to have the file available within the user shares)

 

Myself, even though I wrote that script (and it actually does run on my system), I have all my media shares to not use the cache drive at all.

Squid, you are quite right, that solution would work fine for me. I'm sure your script works fine and I would have used it if it was incorporated into the 'built-in' list and the functionality wasn't so new. My posting was really a way of documenting the problem in case anyone else came across it (it fairly messes up the Plex database!). I guess I was also querying whether the original recommendation to put Dockers on the Cache was still good as without implementation of a solution there is possibly an issue. Finally, I wanted to see whether anyone else had experienced this problem.

 

I ought to add that I had a security camera docker that filled the cache up too, so there are other apps that can cause a problem!

 

Of course using some of the other new CA tools such as Backup etc will also help in this area.

 

I think in my case I am going to go with a slight modification to your solution and that will be to create a transit media fileset where new media files are kept. This will allow me to vet the media (watch/listen) before running a batch archive to the unRaid storage. I have two other reasons for this, one is that I use 'shingled' drives where I believe write once may be advantageous and the second allows me to batch feed new media to a second storage system as an incremental backup.

My strategy is that I have a spinner as my cache drive (2TB as I use it as a PVR) and a SSD mounted outside the array for my docker.img and VMs, my appdata lives on the spinner and I'm perfectly happy with how things are running.