ReadSpeed Result - RAID 5 (4) Samsung 850 EVO 1TB - Question

[Muppet1856]

Interesting result from my run on ReadSpeed. Seems my 3rd SSD (Drive 4) is not playing nice...

How do I correct this? Is it a new drive - Can I run SpinRite on it while not tearing apart my RAID?

Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  120GB OCZ-VERTEX3 MI                362.5   378.8   378.9   378.8   378.8
 82  1.0TB Samsung SSD 850 EVO 1TB       473.8   476.4   474.4   494.1   478.2
  3  1.0TB Samsung SSD 850 EVO 1TB       469.8   472.9   474.6   477.9   479.4
  4  1.0TB Samsung SSD 850 EVO 1TB       287.3    69.7   104.6   526.7   199.5
  5  1.0TB Samsung SSD 850 EVO 1TB       468.2   472.1   474.8   476.7   479.0

                  Benchmarked: Saturday, 2021-01-02 at 13:49

 

[MacNala]

Previous question like yours has been answered by @Steve in the GRC newgroups: Yes

 

[rfrazier]

@Muppet1856 I'm not a RAID expert. But I THINK you would have to set the BIOS so each drive can be addressed individually. Afterward, you'd probably have to set it back to RAID. If you have UEFI, I don't know about that. You could run SpinRite Level 3 on it. That would use up one entire write cycle on the SSD, which reduces its life somewhat. I've seen numbers from 1/300 th of the lifespan to 1/1200 th of the lifespan for the reduction on different drives. That doesn't mean that the numbers couldn't be smaller or bigger. You could potentially run a Level 1 on it first, and make sure the entire drive is readable without problems. If you get read errors, I'd seek more advice here before doing the Level 3, since I don't know how correcting errors would affect the RAID.

Previous question like yours has been answered by @Steve in the GRC newgroups: Yes

@MacNala Many members here, including myself, on the forums are not on the newsgroups.

May your bits be stable and your interfaces be fast. Ron

 

[MacNala]

@Muppet1856 I'm not a RAID expert. But I THINK you would have to set the BIOS so each drive can be addressed individually. Afterward, you'd probably have to set it back to RAID. If you have UEFI, I don't know about that. You could run SpinRite Level 3 on it. That would use up one entire write cycle on the SSD, which reduces its life somewhat. I've seen numbers from 1/300 th of the lifespan to 1/1200 th of the lifespan for the reduction on different drives. That doesn't mean that the numbers couldn't be smaller or bigger. You could potentially run a Level 1 on it first, and make sure the entire drive is readable without problems. If you get read errors, I'd seek more advice here before doing the Level 3, since I don't know how correcting errors would affect the RAID.


@MacNala Many members here, including myself, on the forums are not on the newsgroups.

May your bits be stable and your interfaces be fast. Ron

@rfrazier I am sorry to hear that you have ignored or been unable to view the GRC newsgroups. If you examine the results from my ReadSpeed tests you will note that the drives are individually tested even within a RAID.

 

[rfrazier]

@MacNala Thanks for the info. The more info we can collect and share, the better. I suspected that ReadSpeed would look at the drives individually. But, I didn't know what would happen if the BIOS was set to RAID mode during the test. I also didn't know what would happen to the RAID array if SpinRite Level 2 or 3 were to actually do error corrections on one drive from the array. I've never had any personal experience with RAID systems.

May your bits be stable and your interfaces be fast. Ron

 

[MacNala]

@MacNala Thanks for the info. The more info we can collect and share, the better. I suspected that ReadSpeed would look at the drives individually. But, I didn't know what would happen if the BIOS was set to RAID mode during the test. I also didn't know what would happen to the RAID array if SpinRite Level 2 or 3 were to actually do error corrections on one drive from the array. I've never had any personal experience with RAID systems.

May your bits be stable and your interfaces be fast. Ron

The question about RAIDs other than 1 is important when the drive tests under SpinRite modify bits is a good one. I would hope that Steve will be able to answer. With the RAID in mirror mode both drives should be identical. and device checking by the RAID driver should be detecting that and putting the two drives into Verify mode thus resolving the discrepancy. I have not studied RAID theory and my answers should be treated as unproven.

 

[PHolder]

The question about RAIDs other than 1 is important when the drive tests under SpinRite modify bits is a good one

Well, logically, all drives look to a system as a series of blocks of data (sectors) that can be address by a sequential address (the LBA.) So if SpinRite revives a disk "in passing" and the drive does all the work in the back end, the same "refreshed" data should end up at the same LBA... and the RAID would be non the wiser.

Now if the sector is marked bad I don't know if the RAID will be a fan of that. The other side of the coin is that it wasn't going to be happy anyway.

 

[Barry Wallis]

...
Now if the sector is marked bad I don't know if the RAID will be a fan of that. The other side of the coin is that it wasn't going to be happy anyway.

It seems to me that the disk firmware can decide to mark a sector bad whenever it needs to. RAID magic has to be able to account for that or else it wouldn't work. Am I missing something?

 

[PHolder]

disk firmware can decide to mark a sector bad whenever it needs to. RAID magic

Well in theory there are drives with NAS friendly firmware (see WD RED drives) that turn off this behaviour and just report to the RAID controller to let it decide the next step. I have heard horror stories of NASes going offline on a single bad read when the drive didn't have this kind of firmware. I think there truly is some [bad] black magic back there somehow.

EDIT to add link: https://shop.westerndigital.com/solutions/network-attached-storage/nasware

 

[Barry Wallis]

@PHolder: Thanks for the clarification.