SoftRAID 8 and RAID 4 external Arrays

My previous issue resolved by @SoftRAID is discussed here. This seems like a reversion to me. But waiting to hear further... 

@henry-in-florida 

Encryption is complicated. We are working on this and the reason it is not in the SoftRAID user interface is we have to resolve these issues.

We have support issues filed with Apple on this, its a matter of time.
BTW: if you had noticed, the 8.0 driver was already installed in Sonoma 14.4, when you were running SoftRAID 7.6. You cannot change the SoftRAID driver version, it is determined by MacOS.

This is now an unencrypted disk that had improved performance when left in that state for 24 hours as you suggested. However, after a longer period and after installing the new driver with the new application (8.0), it's now back to being slow again. So, how is this reversion in performance related to encryption?

@henry-in-florida 

My guess is fragmentation of files. APFS is good enough at this, add in the decrypted volume and data is probably scattered to the winds.

Posted by: @softraid-support

↑

@henry-in-florida 

My guess is fragmentation of files. APFS is good enough at this, add in the decrypted volume and data is probably scattered to the winds.

@SoftRAID So you're saying that erase and restore would now be the best way to return it to some semblance of speed? Backing it up to an external device, erase and reformat the RAID and restore the data? Any other way to avoid days'loss of use. Thank god the larger array isn't so badly affected. REALLY regretting the encryption now. 

@henry-in-florida 

I do not have enough experience to recommend anything else, no. Sorry for this.

I guess restoring is the only way.

Posted by: @softraid-support

↑

@henry-in-florida 

I do not have enough experience to recommend anything else, no. Sorry for this.

I guess restoring is the only way.

@SoftRAID Hindsight being what it is, should've tested without encryption first. I did test the array before updating, but did not test after upgrading, before encrypting. Cart before horse, apparently. Hope somebody learns from this. Lots of moving parts. Up to 40% (not 60+ write speed reduction is what I hear.

On an external drive with existing 7.6 driver, the driver seems to remain as-is until you update the app. At least that's what happened to me. While as you said, the internal drive and all new arrays added becomes updated to the new driver. Do I have that correctly?

Encrypting and then Decrypting did several things that were bad:

• Takes a long time both ways. 
• Encrypting the data en situ was REALLY long.
• The shortest way would've been encrypt an empty array>restore the data>install the new driver/RAID 8.0 app.

@henry-in-florida 

Remember as of 13.3, there are no driver updates, that all is controlled by MacOS.

And SoftRAID disks/volumes do not care what driver loads. Unless we make a major change to the SoftRAID disk partition maps, all 8/7/6/5 volumes are forward/backwards compatible with each other. Nothing to worry about.

When we finally add for example RAID 6, old drivers will not mount those volumes. But otherwise, everything is compatible with every other version of SoftRAID.

Encryption is slow, I agree.

Posted by: @softraid-support

↑

@henry-in-florida 

Remember as of 13.3, there are no driver updates, that all is controlled by MacOS.

And SoftRAID disks/volumes do not care what driver loads. Unless we make a major change to the SoftRAID disk partition maps, all 8/7/6/5 volumes are forward/backwards compatible with each other. Nothing to worry about.

When we finally add for example RAID 6, old drivers will not mount those volumes. But otherwise, everything is compatible with every other version of SoftRAID. 

Encryption is slow, I agree.

@SoftRAID In reviewing the screenshots I sent you and repeating them, noticing a high error rate on R/W readings (you'll see it in the screenshots). Also found a flaky cable which shows cutting out at any slight wiggling of one end. I proved it's not the port connector on the machine/hub by swapping ports, so before I do anything else, will try a new 40Gbs 6'certified cable to that array. I also notice that the first time I test with AJA Lite, the speed is much more normal (and without errors), but second and later repeats the speed goes WAAAY down. So I'm suspicious. 

Seeing the same high error rate on single drives in the same enclosure in the cable-fed daisy chain. Moreover, as the error rate goes down the speed goes up, AND vice versa!

Continuing to test and compare with other drives to narrow down the path that run slowly after 3+ iterations in AJA System Test Lite.

So far, finding all drives (not just RAID Volumes) on the 6' existing T4 rated cable (can't tell if it's certified, probably not) and daisy chains with or without a screen connected tend to develop serious errors in R/W that they do NOT exhibit in Read mode. That error apparently is causing a slowdown in performance intermittent during those iterations. AJA has a valuable tool which replicates what I'm also seeing in Activity monitor > Disk performance (confirmation).  The other (6TB) RAID is on a separate cable attached to the same hub and shows no such errors.

Gently wiggling one end of the suspect cable while it's connected and the app is running seems to cause the error rates to increase/get worse, while wiggling the other end does not have the same effect... 

Ended up replacing the T4 cable (with 2.0M OWC) from the hub to the drives. Made a reasonable difference, IMO. PIC. More importantly, it's now reliable and consistent over several iterations and the graphed error rate has come way down (although it varies with the run over frame counts, by observation)! Cables do matter. 

@henry-in-florida 

Where did you get your prior cable from? Did it have a Thunderbolt logo on the end? With a number 3/4 also?

Thunderbolt cables are supposed to be "certified, unlike USB cables, where total junk can be sold.

Posted by: @softraid-support

↑

@henry-in-florida 

Where did you get your prior cable from? Did it have a Thunderbolt logo on the end? With a number 3/4 also?

Thunderbolt cables are supposed to be "certified, unlike USB cables, where total junk can be sold.

@SoftRAID

Belkin. Need I say more? It has the lightning bolt on it and the name and doesn't carry power fairly small diameter. So it should've worked. But... It's Belkin. 

@henry-in-florida 

Makes sense. Belkin went from being (in my opinion) a quality dependable brand, to inconsistent/borderline, over the years.

Thanks!