I am running 6.2.4 with 14 data disks, 2 parity disks and 2 cache disks using 3 AOC-SASLP-MV8 cards

 

On running a monthly correcting parity check, Disk 14 (Data) was red balled due a read error. So I thought that I would stop the array and unassign it, restart the array and this would then allow me a remove the disk to run some tests on it.

 

However when I stopped the array, I was informed that 4 disks were not availible, including Disk 14, which I recognised were all served from one port of one of the AOC-SASLP-MV8 cards, so I thought perhaps a wiring / card problem.

 

I then powered down the array, stupidly not downloading a diagnostics log at this time!, and checked that the 4 disks were connected properly and that the associated card was properly seated too.

 

I then rebooted and all that was showed was that Disk 14 had been reballed, the other 3 disks were now showing OK.

 

So I stopped the array and unassigned Disk 14, and restarted the array. Thereafter I stopped the array reassigned Disk 14, restarted the array and let it rebuild.

 

The array is now up and running OK, however whilst Disk 14 was rebuilding I had a report that I had a pending sector on one of the cache drives?

 

I have run short and extended SMART reports on the cache drive, however they both report back that no error was found. As such I don't know want to do next, I am tempted to run the drive through WD's HD test software - Would writting zeros to the drive force it replace the pending sector with a spare one?

 

As a side note, some months ago I had 2 data disks show red balls, which ultimately proved to be a red herring, and I recall that a member who helped me pointed to a HBA card issue at the time. Afterwards I moved these 2 disks + 2 others (which were on a same card port) onto a spare port on one of the HBA cards (I am only using 5 ports out of 6 availible over 3 cards) and have had no problems since.

 

However I have read on the forum that people have had problems with Marvell controllers, especially where virtulisation is used, but thought that this is mainly AOC-SAS2LP-MV8 cards though? I also understand that a fix has now been applied since 6.1.x?

 

I have now decided to replace the cache drive to be sure, it's only SATA 250GB so cheap!

 

 

 

 

 

When a disk passes the extended SMART test with pending sectors it means that those are false positives, in my experience happens more with WD, there's no way to make them go to 0, disk can be used normally, although that's never a good sign and that disk is more likely to have more issues in the future.

Thanks for your reply, I wondered if it was a false indication too. A replacement disk is cheap as it's only 250GB SATA so better safe than sorry, I have now ordered a replacement

 

I have ready the forum posts at length and see that there was previously problems with Marvell drivers, especially with vurtilisation, but I understand that these issues have been fixed by 6.2.4. I just wonder why this seems to be a 2nd time that I have had. issue relating to a HBA

 

Just to try and rule out a faulty card I have just brought another card from eBay