How to guarantee array uptime

[sjp770]

So the other day I came across the unmountable drive issue. Easily fixed with a bit of time and a forum search, but what I really want to know now is how can I stop the array from not starting if that ever happens again?

 

If there was a way to keep the cache disk live for an isolated from the array appdata share so that my docker apps dont crash i'd be kind of happy.. I know its not true raid, but having a disk fail to mount shouldnt stop the server from running if there is valid parity. Even if all changes were redirected to a temporary location to maintain parity while you fix the root cause - I just need my services to be running at all times.

 

Any tips or do I have the wrong product for any type of higher availability?

[SSD]

UnRaid is prepared to operate with one disk down, even two disks down with dual parity, but it is not able to operate with the array offline yet cache active and dockers running.

 

If you have a good server with solid cable management, drive cages, perform routine parity checks, and check the smart attributes from time to time, you will enjoy good uptime, and avoid an actual drive failure most of the time by replacing a drive when it starts showing signs of failure.

 

It is possible to have a server with a minimal "array" component (one unprotected disk), and your dockers and VMs could happily do their thing on the cache drive without risk of drive causing problems. Then you could have the "real" array on a separate box.

 

It is a valid question though. It would be very nice to be able to have a backup capability to allow Dockers that need to be up all the time to run continue to run in the event of a server issue that takes the primary server down (including hardware issues and issues requiring extensive troubleshooting). I could see a backup machine primed and ready for a cache drive transplant (might have to adjust some template mappings), to allow Dockers to keep working with the primary server down. Might not even need to be primed - allocating a backup machine to this in a pinch could be done pretty easily.