It has been the third time now that disk1 of my server was disabled (Server 1, the one with the Supermicro board). What I did until now:

• checked disk1 (smart report attached)
  
• detached/attached all SATA cables
  
• pulled/reinstalled the SATA card
  
• bought new SAS --> SATA cables
  
• rebuild disk1 two times and all went well. After a couple of days disk1 is "disabled" again or even the rebuild process is breaking up (well, that's clearly because of the REISERFSCK errors)

 

The syslog as of yesterday is attached, the log today is showing loads of these messages:

Aug  7 09:43:23 Tower kernel: REISERFS error (device md1): vs-5150 search_by_key: invalid format found in block 546144281. Fsck?
Aug  7 09:43:23 Tower kernel: REISERFS warning: reiserfs-5090 is_tree_node: node level 46725 does not match to the expected one 1

 

I'm actually running a reiserfsck --fix-fixable /dev/md1 now but I truly believe that the root cause is sitting somewhere else. There are 2 items that could be the culprits...the power supply and/or the AOC-SAS2LP-MV8.

 

The server has installed a Seasonic S12II-330Bronze power supply and I'm not sure if this is a single rail power supply. It also has only 4 SATA connectors (I would need 9), hence I have a lot of adapters installed and the question is if the S12II-330 has enough horse power with it's 330W.

 

I thought to buy one of these as I don't think that I need much more power. My intension is to rather swap smaller against larger drives over time than installing additional drives:

• Seasonic Platinum-400 Fanless (SS-400FL2 Active PFC F3) ...also only 6xSATA but additional 5x peripheral 4 pin connectors
  
• Seasonic S12G 450W (SSR-450RT Active PFC F3) 8xSATA

 

What do you think? Any further advise?

According to Newegg your current 330W PSU is 2 rail.

 

Thanks trurl. That might be an issue as Limetech is very clear: "it’s very important to use a single +12V rail design". However do you (or anyone else) think that this can cause this strange behavior?

 

Interesting news, the reiserfsck ended w/o an error:

root@Tower:~# reiserfsck --fix-fixable /dev/md1
reiserfsck 3.6.24

Will check consistency of the filesystem on /dev/md1
and will fix what can be fixed without --rebuild-tree
Will put log info to 'stdout'

Do you want to run this program?[N/Yes] (note need to type Yes if you do):Yes
###########
reiserfsck --fix-fixable started at Thu Aug  7 13:02:07 2014
###########
Replaying journal: Done.
Reiserfs journal '/dev/md1' in blocks [18..8211]: 0 transactions replayed
Checking internal tree.. finished                                
Comparing bitmaps..finished
Checking Semantic tree:
finished                                                                       
No corruptions found
There are on the filesystem:
Leaves 188658
Internal nodes 1132
Directories 24
Other files 89440
Data block pointers 173316441 (28 of them are zero)
Safe links 0
###########
reiserfsck finished at Thu Aug  7 14:28:29 2014
###########

 

After a data rebuild try the syslog is showing the following:

Aug 7 15:50:01 Tower kernel: md: disk1 write error, sector=26600
Aug 7 15:50:01 Tower kernel: md: disk1 write error, sector=26608
Aug 7 15:50:01 Tower kernel: md: disk1 write error, sector=26616
Aug 7 15:50:01 Tower kernel: md: disk1 write error, sector=26624
Aug 7 15:50:01 Tower kernel: md: disk1 write error, sector=26632
Aug 7 15:50:01 Tower kernel: md: recovery thread sync completion status: -4
Aug 7 15:50:01 Tower kernel: md: recovery thread woken up ...
Aug 7 15:50:01 Tower kernel: md: recovery thread has nothing to resync
Aug 7 15:50:36 Tower kernel: mvsas 0000:02:00.0: Phy2 : No sig fis
Aug 7 15:50:39 Tower kernel: sas: sas_form_port: phy2 belongs to port6 already(1)!
Aug 7 16:00:01 Tower logger: mover started
Aug 7 16:00:01 Tower logger: skipping */
Aug 7 16:00:01 Tower logger: mover finished
Aug 7 16:03:41 Tower kernel: mdcmd (46): spindown 0
Aug 7 16:05:41 Tower kernel: mdcmd (47): spindown 2
Aug 7 16:05:42 Tower kernel: mdcmd (48): spindown 6
Aug 7 16:05:42 Tower kernel: mdcmd (49): spindown 7

 

So should I change the power supply as next step? Does it make sense? Is there anything that someone could read from the syslog/smart report?

 

Thank you.

The PSU is insufficient.

According to Newegg your current 330W PSU is 2 rail.

 

Most PSUs that claim to have 2 (or more) rails are not (at least not the way you might be thinking).

http://www.jonnyguru.com/modules.php?name=NDFAQs&op=FAQ_Question&ndfaq_id=20'>http://www.jonnyguru.com/modules.php?name=NDFAQs&op=FAQ_Question&ndfaq_id=20

and

http://www.jonnyguru.com/modules.php?name=NDFAQs (section about multiple rail near the bottom)

There is also this Seasonic page which shows 2 rails with 17A each.

There is also this Seasonic page which shows 2 rails with 17A each.

That's the kind of thing I mean, likely all the +12v are tied together internally. 

 

On another note, I don't know if this is is a valid test, but in a multi-rail PSU should all the +12 (yellow) wires have continuity between each other (even if they are on separate rails)?

Dear all,

 

thanks a lot for looking into this. Seems that you have all strong opinions about the actual power supply. The calculation that I made when I bought the PSU: 10W per disk drive and 100W for the rest of the HW. The point that is making me suspicious: why does the issues is coming up now?

 

What PSU should I buy? The Seasonic S12G 450W (SSR-450RT) with 8xSATA or the Corsair RM450? Or is that wattage still not sufficient? In that case I would go for the Platinum-660 (SS-660XP2) - all of these seem to be all single rail PSU's.

 

Last question: have you seen anything odd in the log and/or in the smart report? The disk is ok? I couldn't saw anything but here are much smarter people than me...

 

Thank you.

try swapping the power adapter/splitter from the problematic disk to another and see if the problem follows the power adapter/splitter.

 

One of the moderators on here had an issue causing by a bad power splitter.

Good idea mr-hexen but I should have mentioned it at the very beginning that I have tried that before. But thanks for reminding...

 

I ordered a new PSU: the Seasonic Platinum-660 (SS-660XP2). Should have enough plenty of horse power and has a single +12V rail.

 

But let me finally ask the question once gain: has anyone seen anything odd in the log and/or in the smart report? The disk is ok?

 

Thanks a lot.

Sorry, didn't mean to get into a PSU war.  As far as what it's "coming up now", in my experience when PSUs fail pretty much any kind of bizarre behavior can ensue.  If you're lucky the PSU will just flat-out die and the problem will be obvious, unfortunately other times things just start behaving badly for no apparent reason.  I just recently had the PSU in my unraid go bad and the symptom was that one specific disk would either redball or fail to be recognized by the controller at boot (intermittently) or it would boot and start fine (all disks spinning but not necessarily reading) and they if I start a parity check (all drives reading) I would get errors all over the place.  The PSU in question was probably a poor choice but it worked completely fine for quite some time before it gave out.

 

The power supply you chose seems like it's very well made, if you are interested (and haven't seen it) there is a well done review at http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=326

 

I especially like that they take apart and examine the PSUs and that build quality is part of their scoring methodology (they're not the only ones that do this, but among the first).

No need to apologize WizADSL - I appreciate any advise and your experience seems to be VERY similar to that what I made.

 

And thanks for the link - I saw one German review, but this is indeed new news to me

 

Cheers

this needs to be resolved:

 

Aug  6 09:05:43 Tower kernel: FAT-fs (sda1): Volume was not properly unmounted. Some data may be corrupt. Please run fsck.

 

 

This comes up every time. However I'm going to pull the flash and run a chfdsk/f after having installed the PSU. Thanks for having looked into the syslog.

I have changed the PSU now, but after all it seems that I have to RMA the disk drive:

root@Tower:/boot# smartctl -a -A /dev/sdi
smartctl 6.2 2013-07-26 r3841 [x86_64-linux-3.10.24p-unRAID] (local build)
Copyright (C) 2002-13, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF INFORMATION SECTION ===
Vendor:               
Product:              F?
Revision:             ?
scsiModePageOffset: response length too short, resp_len=47 offset=50 bd_len=46
>> Terminate command early due to bad response to IEC mode page
A mandatory SMART command failed: exiting. To continue, add one or more '-T permissive' options.

 

There is one thing I wanted to as k as well. Is that syslog message important:

Tower kernel: program smartctl is using a deprecated SCSI ioctl, please convert it to SG_IO

I have a new PSU now, the cabling is brand new but the real issue was the drive. I have changed it now and everything's fine. The primary scepticism came because it's the second time that md0 was the culprit but it turned out that this was just by coincidence. This was the reason that I thought that the server has another faulty part.

Thanks to all who were helping here.