Hi guys,

 

I have run into a bit of an issue trying to pre-clear a new drive. The drive in question is a WD EARS.

 

When I run the the pre-clear command as follows:

 

preclear_disk.sh -A -c6 /dev/sdg

 

The pre-read starts, but crashes after a varying period of time (once after 10 minutes, a second time after 2 1/2 hours). I can't access the array after the pre-clear has crashed, so have to run the 'shut-down' command to reboot.

 

I have run the Smart report, and can't see any issues with the drive. (Copy attached unless I'm missing something).

 

I also attach a Syslog, captured after the pre-clear had crashed, which seems to show lots of errors relating to the drive in question (sdg) - but unfortunately its all far beyond my understanding!

 

Any help anyone can give would be greatly appreciated.

 

smart.txt

2011-03-23_Syslog.zip

I didn't go through your entire syslog (it was really long!) but I saw a lot of Buffer I/O errors towards the beginning.  It could be a bad cable (data or power).  First try swapping out the cables, or trying the drive in another (known good) drive slot/bay.

 

Are you running preclear from the system console or telnet?  If telnet, are you using screen?

I didn't go through your entire syslog (it was really long!) but I saw a lot of Buffer I/O errors towards the beginning.  It could be a bad cable (data or power).  First try swapping out the cables, or trying the drive in another (known good) drive slot/bay.

 

Are you running preclear from the system console or telnet?  If telnet, are you using screen?

 

Hi Rajahal,

 

Preclear is running from the system console. I will try the drive in a different bay. I am using IcyDock hotswap backplane bay, connected up to a 3Ware AOC-SASLP-MV8 card via breakout cables. Other drives are working fine in the same Icydock 'box', connected by the same cable to the same card.

 

Assuming the drive works in one of the other bays, would the most likely cause be a faulty power cable? (I am using an Enermax 1000W power supply)

I didn't go through your entire syslog (it was really long!) but I saw a lot of Buffer I/O errors towards the beginning.  It could be a bad cable (data or power).  First try swapping out the cables, or trying the drive in another (known good) drive slot/bay.

 

Are you running preclear from the system console or telnet?  If telnet, are you using screen?

 

Hi Rajahal,

 

Preclear is running from the system console. I will try the drive in a different bay. I am using IcyDock hotswap backplane bay, connected up to a 3Ware AOC-SASLP-MV8 card via breakout cables. Other drives are working fine in the same Icydock 'box', connected by the same cable to the same card.

 

Assuming the drive works in one of the other bays, would the most likely cause be a faulty power cable? (I am using an Enermax 1000W power supply)

You may have a 1000 watt supply, but it is a multi-rail supply with 5 24amp rails.

 

You have 12 disks in your server, and it is highly likely they are all on the same 12 volt rail.  That rail might even be shared with the motherboard and all the case fans.  It is probably getting very close (if not over) its current rating, especially if any of the disks are not low-power green" drives.  I'm mentioning this only because you might want to look into how that power supply uses the capacity it advertises.  It might be you have less usable capacity than some single rail 500 watt supplies

Hi Joe,

 

Thanks for your reply. I will look into this. The drives are all green drives I think, with the exception of a couple of older 250Gb drives. Is there any way I can utilise one of the other rails then to power one of the Icy Docks? Or do I need to look at getting a new power supply?

 

Looking at the syslog, do you think the problem is most likely to be a power one?

Checked my PSU and it is infact a 600W version (the 1000W unit is in another gaming rig I think).

 

The PSU in my Unraid server is in fact an older 600W Enermax Noisetaker. It does indeed have seperate 12v rails, each only providing 18A!!! 

 

I'm guessing I need a new PSU??!! How big should I go if I want to cover future expansion up to the maximum 22 drive array? How many amps should I allow per 'green' drive, and how much for a full fat 7200rpm drive?

 

Looking through the forum, the Corsair models seem to be a favourite, would a 750W Professional (HX750) be sufficient - giving 62A on a single 12v rail?

 

Thanks in advance...

yes, that would be enough for a mixture of green and non-green drives.

 

in general assume 2amps for a green drive and 3 amps for a non-green drive

Does the MB use some of the 12V amperage?

Does the MB use some of the 12V amperage?

yes, and so do the case fans and disk controller cards.

@ Prostuff1 - Thanks for that, sso for absolute future proofing, should I step up to the 850W version (70A over a single rail) for the extra £15 it would cost - since that would cover the option of 22 x 7200rpm drives (3amp x 22 = 66amp)?

 

@ dgaschk - I'm not sure, and I'm not sure quite how to find out?? It is a BFG nVidia 680i SLI board out of an old gaming PC. I'm also running it with a graphics card (since it has no on board graphics) so I can access the console - this requires a molex connector to be plugged into the MB to supplement the dedicated graphics card power cable, so perhaps this uses the 12v as well?  

 

@ Joe. L - Based on these other overheads, how big a PSU should I be going for? Is the HX850 even enough??

@ Joe. L - Based on these other overheads, how big a PSU should I be going for? Is the HX850 even enough??

The CPUs and graphics cards use 12Volt power too.

 

I think the 850 watt single rail supply you mentioned will be more then enough.  Odds are highly likely you'll have more "green" drives as you grow in the future than non-green.

Checked my PSU and it is infact a 600W version (the 1000W unit is in another gaming rig I think).

 

The PSU in my Unraid server is in fact an older 600W Enermax Noisetaker. It does indeed have seperate 12v rails, each only providing 18A!!!  

 

I'm guessing I need a new PSU??!! How big should I go if I want to cover future expansion up to the maximum 22 drive array? How many amps should I allow per 'green' drive, and how much for a full fat 7200rpm drive?

 

Looking through the forum, the Corsair models seem to be a favourite, would a 750W Professional (HX750) be sufficient - giving 62A on a single 12v rail?

 

Thanks in advance...

 

I have the non-modular model (TX750), and it has been great (currently on 12 drives). Also, you can save a few bucks if modular isn't absolutely necessary. Here's a CDN vendor that has it on sale for CDN$90:

 

http://www.ncix.com/products/?sku=26415&affiliateid=7120027

 

Here's the US version (same price):

 

http://www.ncixus.com/products/?sku=26415&vpn=CMPSU-750TX&manufacture=Corsair&promoid=1302

 

At 2amps a pop, 20 green drives is still only 66% capacity, amp-wise (still leaving room for other bits using 12v)...

 

[edit] Just saw the "£" sign, so my links may be of no use . If the 850 is only a bit more, then might be worth it....

Before you start replacing components, I suggest you shut down the other drives and only preclear the new drive. If it works, then could be a weak psu, if it doesn't work then I'd suggest trying preclear again in a different slot to see if its th drive or the controller/cable.

 

I had similar issues trying to preclear 6 new drives, turns out 1 of the new drives was bad and caused syslog explosion which crashed the machie, makig the other drives look bad too.

Right, just a bit of an update on this issue:

 

I tried the drive in a different working drive slot, and still the same crash, so I have replaced the power supply for a shiny new Corsair AX850 which provides 70A on a single rail as recommended above. I know this is a little bit overkill, but I wanted to cover all future possibilities in one hit.

 

I re-ran the pre-clear script and still no joy, pre-clear crashed again part way through the initial read. I have even tried running pre-clear by skipping the initial read, it gets all the way through the write but crashes on the post-read.

 

I also checked the slot the drive was in by pre-clearing a different old, known working drive in it, and that pre-cleared without a problem!??

 

Anyway, I have now sent the drive back to Scan for a replacement.

 

Does anyone have any suggestion of what might have been wrong with this drive? Why is the Smart report not showing any errors?

Unfortunately broken drives don't always follow the rules and report errors in the typical ways. Your experience here is not typical. You did a nice job isolating the problem to that disk, so hopefully the replacement will work properly.

Does anyone have any suggestion of what might have been wrong with this drive? Why is the Smart report not showing any errors?

 

Bad shipping perhaps - it has a multizone errors which are often associated with the mechanical worthiness of the drive.

hmm... sounds like the problem I had. out of 6 new drives, 1 was bad and would cause log explosion which ultimately crashed the system.

This is food for thought!

 

I had a couple of bad EARS drives that failed on pre-clearing and would  crash the system.  I didn't know what was doing it (I just assumed it was bad drives) at the time, but exploding log file makes sense... Especially since I added 2 more GBs of ram (4GB total) and went down to preclearing one drive at a time, and they still failed the pre-clear, but they completed without crashing...

Right, so what are you thinking? That the systems trying to make so many writes to the log file that it causes a system crash? My system only has 2GB but I thought the UnRAID system wasn't very demanding in terms of hardware requirements?

Right, so what are you thinking? That the systems trying to make so many writes to the log file that it causes a system crash? My system only has 2GB but I thought the UnRAID system wasn't very demanding in terms of hardware requirements?

 

It is not, but if you have a drive that is not playing nice the syslog can quickly grow out of control and to large, causing the server to crash.

My system only has 2GB but I thought the UnRAID system wasn't very demanding in terms of hardware requirements?

 

Keep in mind, I don't have any way of proving that my problem was the same as being discussed, but the symptoms sound very similar... Furthermore, the fact that when I doubled the ram and reduced the number of drives being simultaneously pre-cleared (and by so doing, reducing the number of log writes), the system stopped crashing, makes me believe that they were the same issue...

 

As far as the demanding-ness of unRaid, I wouldn't worry about 2GB of ram. Before the crashing, 2GB was more then enough for me. I pre-cleared multiple drives before without issue, and haven't seen any measurable performance improvements from the extra ram in everyday use (Other then it stopped the server from crashing while pre-clearing the bad drive). That said, I have left it in, simply because it's there now and I have no other pressing need for it (Well, I did pull it from my gf'S Hackintosh, but it seems happy with just 2GB...  )

just food for thought... when i had my new drive fail, it caused the syslog to grow to beyond 2gigs... which ultimately crashed my system.. that is when I moved the syslog to a tmpfs instead to limit its impact on the server.

just food for thought... when i had my new drive fail, it caused the syslog to grow to beyond 2gigs... which ultimately crashed my system.. that is when I moved the syslog to a tmpfs instead to limit its impact on the server.

 

Perhaps that should be a default setting to avoid this issue permanently. What are the cons to moving the syslog to a tmpfs?

You'll have to excuse my newbie ignorance, but what is a tmpfs?

just food for thought... when i had my new drive fail, it caused the syslog to grow to beyond 2gigs... which ultimately crashed my system.. that is when I moved the syslog to a tmpfs instead to limit its impact on the server.

 

This could have been the problem with mine too as my syslog got to about 1.7GB I think