4 minutes ago, guru69 said:

Am I right in assuming RAID10 will provide the best write speeds/reliability for the cache pool?

It would be the best option for both speed and redundancy, if you can use 4 devices.

OK. I will order one more SSD, Thanks!

Thought, a 3 NVMe raid1 pool might perform better than a 4 SSD raid10 pool, depending on the devices used.

What do you recommend to test read/write speed on the cache RAID?

I just ordered 4th SSD, but am interested now in testing both ways.

11 hours ago, guru69 said:

What do you recommend to test read/write speed on the cache RAID?

I usually prefer testing in real world situations, i.e., by copying/reading files to the pool, but if you're more interest in raw speeds you can use dd, google dd speed test.

Yes I've used DD in the past, mostly interested in raw speeds. I wasn't sure if there was another tool or plugin in Unraid that was better. I will post my results to forum,

Thanks for your advice

Here are my results from DD...

 

(Before)

2 x Samsung 850 Pro 256GB Sata3 in Raid1 -> 425MB/s to 465MB/s

 

(After)

4 x Samsung 960 Pro NVME in RAID10 -> 1.5GB/s to 1.6GB/s

 

I'm a happy camper ?

 

Maybe I spoke too soon. After 24 hours w/ 4 SSDs on RAID10, Got a message about a cache drive missing. Logs showed lots of "read errors corrected.." and 2 of my dockers disappeared. I ran the Preclear/Erase all Data on the newly unassigned SSD, all 100% Success. I switched the SSD drives around to different sockets (new build, to rule out the socket). I attempted once more on RAID10 with same result (on new socket, same SSD). I switched the array to RAID1, all errors in log have ceased, so will restore appdata and see how it goes.

 

Interestingly, DD reports the exact same speeds of 1.5GB/s to 1.6GB/s using RAID1, hopefuly its a bit more reliable.

 

40 minutes ago, guru69 said:

Got a message about a cache drive missing

Device dropping won't have anything to do with raid mode, it's a hardware/connection problem.

If it were a connection issue I'd assume once it landed in Unassigned Devices that it would produce errors there as well. It completed Preclear twice and passed Smart tests. It also produced the same odd behavior in a different socket once back in the cache pool. I haven't seen any issues on RAID1 yet (up for an hour now), but I'll keep you posted. If its a connection issue then I'm sure the errors will return (fingers crossed).

guru69 - Are you using the 4 SSD's in RAID1 or 2 of them? Any update on the errors you were getting?