Raid 5 failure with TWO drives down ARECA

W

WoooT

n00b
Joined
Aug 1, 2010
Messages
43
Once apon a time there was a raid 5 array with 8*1,5TB drives, one of those drives failed lets call him disk no5. Disk no5 was a hard working 2year old who couldnt work anymore, so we had to replace him. His brand new replacement the seagate 2TB LP took the job as disk no5.

The new disk no5 worked like a charm for 4 month before he was tired of his job so he quit. We where mad at this point so we got another new seagate 2TB to take his brothers place.

This young brother worked 3 days before he got lazy and got fired from his position as disk no5, he was so mad he got fired so he took disk no6 with him he said i wont go down by myself.


This little story is what ive experienced the past 5 months, 3 disks in the same channel5 quit there job, coincidence? or is those LP Seagate drive totally CRAP? i dont really give a fuck atm becuse i now have 2 failed drive and my RAID 5 i degraded you now what that means.

Anyway i fired up my system, and i can now se that i have 2 raid arrays, they are both identical exept one contains disk 1,2,3,4,x,x,7,8 and the other contains x,x,x,x,x,6,x,x

So it seems disk 6 is fine i think it went down because its sharing same power cord as disk5

Is there any chance to merge those two "raids" so it could be fully functional?



Ive read this thread and this guy had same issue as me, seems he got it sorted out, will this work for me?
http://hardforum.com/showpost.php?p=1032176742&postcount=26
 
Seagate desktop drives are aweful, this is proving my point from other threads even more.

Seriously, Seagate has great nearline and enterprise class drives, and even their GoFlex drives are awesome.

However, their desktop drives are horrendous and fail left and right. Every time I have someone come to me with a dead drive it has always been a Seagate desktop (Barracuda) drive.

WoooT, if I were you, I'd move to Samsung or WD for desktop drives and back everything off of your existing array before it is too late. It may be a bit expensive, but it will be up to you if your data is worth it or not.

EDIT: As for that method, I've never tried that before, but honestly I'm not sure if I would trust putting in only two drives considering a lot of hardware RAID cards are flaky and may break the array in the card's firmware settings.
 
Last edited:
Thx for the reply, well yeah those Seagate LP are totally crap. but to buy 2TB raid edition disks cost 3 times more.

isnt there any good reliable consumer drives out there?, from now on i wont go near Low power/green drives any more. i´ll just buy regular ones and hopefully theyll last longer.

Bet before i can replace any drives and need to get the array back on its feet, thats where you guys comes in and tell me how i do recover this mess :).
 
Yeah, the Samsung F3/F4 models are very reliable and work great in hardware RAID, and it doesn't cost an arm and a leg like the WD RE3 drives.
 
Samsung F4 seems like great choice, 100bucks for 2TB but those are LP/Green drives aswell... after what ive been thru im sick of everything called LP nor Green.

Edit just saw the big thread here about Samsung F3/F4 with defective firmware no thanks...
 
with 8 drives, you simply need RAID 6.
anything over 4 I would say and you should go with RAID 6.

I hope you had nothing important on them.
Also, make sure that the drives are not stacked one on top of the other, they need a fan blowing on them for proper cooling.
 
WoooT said:
Samsung F4 seems like great choice, 100bucks for 2TB but those are LP/Green drives aswell... after what ive been thru im sick of everything called LP nor Green.

Edit just saw the big thread here about Samsung F3/F4 with defective firmware no thanks...
Click to expand...

Hitachi may be the better choice then. ;)
 
I recently bought 8 F4EG 2TB drives which have Jan 2011 production date. I ran a bunch of tests on them and concluded that they are not affected by the firmware bug.
 
@OP

Is disk 6 completely screwed? Will it even spin up? Could you try powering it from a separate el cheapo PSU?

omniscence said:
I recently bought 8 F4EG 2TB drives which have Jan 2011 production date. I ran a bunch of tests on them and concluded that they are not affected by the firmware bug.
Click to expand...

That is very good to know. :)
 
parityboy said:
That is very good to know. :)
Click to expand...

Of course, this is not 100% safe, but I ran a 'watch smartctl -a /dev/sdX' on them while copying a 4 TB partition from my old RAID to the new one and the md5sum was the same afterwards.
I also tried the 'mybadblocks' script from the other thread.
 
Last edited:
parityboy said:
@OP

Is disk 6 completely screwed? Will it even spin up? Could you try powering it from a separate el cheapo PSU?
Click to expand...


Yes disk 6 is completely fine, it works from different machine to, it is disk 5 which is screwed...
 
omniscence said:
Of course, this is not 100% safe, but I ran a 'watch smartctl -a /dev/sdX' on them while copying a 4 TB partition from my old RAID to the new one and the md5sum was the same afterwards.
I also tried the 'mybadblocks' script from the other thread.
Click to expand...
Did you enable AHCI and NCQ during this? Without it, the bug would not present itself.

As far as i know the bug is that write requests do not happen while they are said to be completed whenever a SMART abort requests is sent together with a write request, while NCQ is enabled. So two requests at once both a SMART and a write request. But they cannot happen together if NCQ is not enabled.
 
@OP

If disk six is OK, surely then you can add it back to the RAID, which will be running in degraded state, right? If so then yeah, replace disk 5 ASAP with something a little more reliable. I advocate either the Samsung or the Hitachi.

The Samsungs may need to be patched (better safe than sorry), but it's not an onerous task. I have three of them running in Linux MD RAID 5, and they've been great.
 
Well disk 6 is back on its place but it wont join the original raid array...

Ive tried rescue raid, Level2rescue etc, nothing.

is there a way to force disk 6 to join the raid array? Since disk 6 already contains some of the parity and is not broken the array should work?
 
parityboy said:
The Samsungs may need to be patched (better safe than sorry), but it's not an onerous task. I have three of them running in Linux MD RAID 5, and they've been great.
Click to expand...

I would recommend this as well.

The OP may want Hitachi's, but they are no where near as good as Samsung drives, at least at the desktop level drives.

My gf's laptop just had a dead drive, surprisingly it wasn't a seagate, it was a hitachi. ;)

OP, I'm telling you, Samsung have very reliable drives, even if the bug is present, just upgrade the firmware, it goes really fast.
 
sub.mesa said:
Did you enable AHCI and NCQ during this? Without it, the bug would not present itself.

As far as i know the bug is that write requests do not happen while they are said to be completed whenever a SMART abort requests is sent together with a write request, while NCQ is enabled. So two requests at once both a SMART and a write request. But they cannot happen together if NCQ is not enabled.
Click to expand...

I suppose it was enabled, the drives were connected to a 9211-8i, the queue depth was set to 31 (standard) which should imply AHCI and NCQ.

Furthermore there was an article in the last german c't maganzine which states that all F4 drives with a build date of Dec 2010 and later already include the firmware patch.
 
WoooT said:
Well disk 6 is back on its place but it wont join the original raid array...

Ive tried rescue raid, Level2rescue etc, nothing.

is there a way to force disk 6 to join the raid array? Since disk 6 already contains some of the parity and is not broken the array should work?
Click to expand...

Yes it should...but I have no experience of Areca, so I can't give you specifics, but yes as long as the disk has not been modified in anyway, it should rejoin the array with no issue. Could you try looking for something in the card's BIOS, if there is no option in the web management software?
 
parityboy said:
Yes it should...but I have no experience of Areca, so I can't give you specifics, but yes as long as the disk has not been modified in anyway, it should rejoin the array with no issue. Could you try looking for something in the card's BIOS, if there is no option in the web management software?
Click to expand...

oddly enough,The web management seems to have more options than the bios.
 
You must log in or register to reply here.
Back
Top