Parity check neccessary after disk rebuild?

[hawihoney]

After successful disk change for parity or data disks I always did a parity check to prove that everything went fine.

 

Now with the 6TB disks its a week to go for disk_preclear/disk_change/parity_check (old LimeTech Tower from 2007 and 2008). Is this additional parity check step really neccessary or can I trust a successful operation after changing disks?

 

How do you guys do that?

 

Thanks.

 

[sgibbers17]

Is it necessary? No.

 

Should you do one anyways? Probably a good idea.

 

Since you can still use your array during a check I say why not.

[garycase]

I would absolutely do a parity check after you've changed the parity disk and the rebuild is done.  That's the only way to KNOW that all went well.

 

It's also very unlikely that the check will take a week -- it could take that long to pre-clear the disk (not necessary, but a good check of the disk's integrity) and then do the replacement, but the actual check isn't likely to take more than a day, and even that's a long time ... but could be the case if you're using older PCI based SATA controllers.

 

[sgibbers17]

I would absolutely do a parity check after you've changed the parity disk and the rebuild is done.  That's the only way to KNOW that all went well.

 

It's also very unlikely that the check will take a week -- it could take that long to pre-clear the disk (not necessary, but a good check of the disk's integrity) and then do the replacement, but the actual check isn't likely to take more than a day, and even that's a long time ... but could be the case if you're using older PCI based SATA controllers.

I think he means that the hole process from preclear to parity check takes about a week

 

And yes you should do a check.  What I meant by not necessary is that your server is fully usable.  But until there is a parity check your array may not be protected but is most likely is protected but you don't know till the check.

[hawihoney]

Thanks guys.

 

Yes, I'm talking about the complete cycle. I do have two new 6TB drives. The first one, the new parity disk, is in preclear (1 cycle) since 72 hours and waaay to go. Plus parity rebuild, plus parity check - you get the idea.

 

After that the second disk will replace a data disk. Another week to go.

 

This old Tower does not tolerate that much in parallel. Writing few bytes or reading few bytes during parity check bricked the machine several times in the past years. If I do have luck only the Web-UI is gone. Often it locked me out - even SSH didn't work any longer then. So I don't do anything with the Tower during parity build/check - but this has doubled since last years 3TBs and becomes nasty now.

 

On that older Tower (2007) rebuilding a 3TB lasts for 50 hours. On the Tower from 2008 it's down to 32 hours. Would love to change mainboard, CPU, SATA cards - but I don't have any experience with hardware. More time to go outside ;-)

 

[trurl]

I usually have activity (torrents) during parity checks and have never noticed any problems. "Bricking" whatever you mean by that is definitely not expected behavior. There is nothing about the design of unRAID that will prevent using the array during a parity check. The worst that should happen is slower parity checks.

[garycase]

I suspect the issue with your drives causing notable system impact is that you're using PCI SATA cards.

 

If your motherboard has some PCIe slots, you could improve this notably by switching to better controllers -- although unless the slots are PCIe x4 or better that's probably not worth the effort.

 

You may just have to live with that until you're ready to upgrade to a more modern motherboard/CPU combo.