So for a long time (as long as I can remember?) I have been living with slow and inconsistent write performance on my unRaid server.

 

However, as the situation currently sits I am running version 5.0.4 Plus license. I've got 6 HDDs ('green' models) that are all 2TB except for my 4TB parity drive. In addition, I've got an mSata 64GB cache drive that was recently added though a mSata -> PCIe card on my Supermicro Atom motherboard. Right now as I copy a few 2GB files from one location on the unRaid server to another, I'm only getting about 13MB/sec write speeds.

 

I thought this may just be due to my slow green hard drives, which is why I added a cache SSD drive. Initially (maybe for a day or a week?) I would usually get around 50MB/sec write speeds. It is strange because I feel that my read performance is acceptable. ~90MB/sec avg during a parity check, and using a user written script to check HDD read performance I got around 90-100 MB/sec on each drive.

 

Additionally, I recently rebuilt my desktop and started downloading torrents directly to my unRaid (using a standard Windows computer to run the actual torrent program). Write performance has suffered even more, and at times when I'm playing back music from the server, downloading torrents to the server, and copying or moving a file to the server the music will begin to skip. I'm not positive but I think this has to do with the HDDs spinning up.

 

I am currently running no add ons (except for the web interface upgrade) and the unRaid server is 'headless' just controlling through IPMI.

 

Attached is my syslog. Any thoughts?

syslog.txt

and a screen shot of my web interface/info tab

disk3: [8,32] (sdc) WDC_WD20EARS-22MVWB0_WD-WCAZA3972250

Your disk #3 has a problem dating back to early February...there are a number of read failures.

 

Feb  6 04:18:56 Tower kernel: ata2.00: exception Emask 0x0 SAct 0x0 SErr 0x0 action 0x0

Feb  6 04:18:56 Tower kernel: ata2.00: irq_stat 0x40000001

Feb  6 04:18:56 Tower kernel: ata2.00: failed command: READ DMA EXT

Feb  6 04:18:56 Tower kernel: ata2.00: cmd 25/00:00:3f:6c:ea/00:04:e3:00:00/e0 tag 0 dma 524288 in

Feb  6 04:18:56 Tower kernel:          res 51/40:8f:af:6d:ea/00:02:e3:00:00/e0 Emask 0x9 (media error)

Feb  6 04:18:56 Tower kernel: ata2.00: status: { DRDY ERR }

Feb  6 04:18:56 Tower kernel: ata2.00: error: { UNC }

Feb  6 04:18:56 Tower kernel: ata2.00: configured for UDMA/133

Feb  6 04:18:56 Tower kernel: sd 2:0:0:0: [sdc] Unhandled sense code

Feb  6 04:18:56 Tower kernel: sd 2:0:0:0: [sdc] 

Feb  6 04:18:56 Tower kernel: Result: hostbyte=0x00 driverbyte=0x08

Feb  6 04:18:56 Tower kernel: sd 2:0:0:0: [sdc] 

Feb  6 04:18:56 Tower kernel: Sense Key : 0x3 [current] [descriptor]

Feb  6 04:18:56 Tower kernel: Descriptor sense data with sense descriptors (in hex):

Feb  6 04:18:56 Tower kernel:        72 03 11 04 00 00 00 0c 00 0a 80 00 00 00 00 00

Feb  6 04:18:56 Tower kernel:        e3 ea 6d af

Feb  6 04:18:56 Tower kernel: sd 2:0:0:0: [sdc] 

Feb  6 04:18:56 Tower kernel: ASC=0x11 ASCQ=0x4

Feb  6 04:18:56 Tower kernel: sd 2:0:0:0: [sdc] CDB:

Feb  6 04:18:56 Tower kernel: cdb[0]=0x28: 28 00 e3 ea 6c 3f 00 04 00 00

Feb  6 04:18:56 Tower kernel: end_request: I/O error, dev sdc, sector 3823791535

Feb  6 04:18:56 Tower kernel: ata2: EH complete

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791472

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791480

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791488

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791496

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791504

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791512

Feb  6 04:18:56 Tower kernel: md: disk3 read error, sector=3823791520

...

 

Run a SMART test on that drive.

http://lime-technology.com/wiki/index.php/Troubleshooting#Obtaining_a_SMART_report

---

If that comes up clean, run Reiserfsck to check the filesystem on that drive.

http://lime-technology.com/wiki/index.php/Check_Disk_Filesystems

Thanks, SMART report attached. Looks like it is bad. Can this cause poor write performance even to my cache?

smart.txt

SMART report looks okay.

Run Reiserfsck and check the file system.

Instructions are here:

http://lime-technology.com/wiki/index.php/Check_Disk_Filesystems

 

 

I ran SMART again and it does have some errors, in addition to many of the attributes showing "old age" and "pre fail"

 

I assume this is bad? Have not looked into the filesystem check yet

smart.txt

None of the SMART reports look good. See here: http://lime-technology.com/wiki/index.php/Troubleshooting#Resolving_a_Pending_Sector

OK, I have a new HDD preclearing now, will replace in the array later this weekend.

 

Either way, can this really be the cause of my poor performance??

Of course. The system must spinup and read from all disks to reconstruct the data in an unreadable sector after the system gives up on trying to read the bad sector.

Dang. I missed the 38 pending sectors in Smitty2k1's first SMART report. 

Thanks for being here, dgaschk.

Chaos could have ensued...a red face is bad enough.

Sadly, after adding new disk to my system performance is still poor. Updated syslog attached.

 

Thanks for the help so far guys

syslog_updated.txt

Remove the web interface upgrade and update to 5.05.

Same speeds unfortunately.

Attach a new syslog.

Attached, thanks for your continued responses!

syslog.txt

Does unRAID main show any errors?

No, and when I replaced the drive a few days back I ran it through a parity check with zero errors at roughly 90MB/sec.

 

I'm not sure if this is normal, but does it take ~10 seconds for a drive to spin up? Whenever I go to access SMB shares through Windows Explorer, if the drives are spun down everything hangs for about 10 seconds before the window will refresh and the data will be accessible.

I'm not sure if this is normal, but does it take ~10 seconds for a drive to spin up? Whenever I go to access SMB shares through Windows Explorer, if the drives are spun down everything hangs for about 10 seconds before the window will refresh and the data will be accessible.

IN my experience, this is not atypical.

I'm not sure if this is normal, but does it take ~10 seconds for a drive to spin up? Whenever I go to access SMB shares through Windows Explorer, if the drives are spun down everything hangs for about 10 seconds before the window will refresh and the data will be accessible.

IN my experience, this is not atypical.

My experience is similar when the drive is spun down and needs to be spun up. Happens with the 'D:/' drive in my PC, too.  Windows (and Mac) both like to create and paint little custom preview mini-icons on the files and folders in Explorer (Finder). That means all the files have to be accessed even for something as simple as a directory listing. There's a Preference that turns this off on remote drives. (Its in Finder Prefs on a Mac...and there's something similar in Windows 7)

 

In unRAID there are also some 'tricks' that can be used to reduce the frequency of this happening:

• 'Split Level' solves the problem of drive spin up in the middle of your Video playback.
  
• 'Cache directories' removes some of the delays for simple directory accesses. 
  
• 'Spin up groups' can make multiple drives spin up at once. 
  
• And there's a plugin that will cause unRAID to spin up if it sees certain devices as being "ONLINE". (I haven't tried it, but for example, if your Roku box is 'on', the plugin causes unRAID to spin up drives.  Presumably it could be modified for your PC.)
  

But if its 'just' PC access for occasional data files, and you 'know' in advance that you're going to need to access unRAID, you might form the habit of deliberately accessing the unRAID drives early in your PC session, thus spinning up the drives in advance of your frustration. 

Thanks, that all makes sense.

 

I'm still experiencing less than ideal performance. Sometimes just playing back music over the network doesn't work in addition to my slow write speeds. Pretty much out of ideas here

Your syslog indicates that unRAID is running at GigE speed.

and that your ethernet switch autosenses the speed.

If you're having dropouts of music (tunes typically have very low bandwidth), I'd look at the rest of your network.

For example, if you connect your PC and unRAID via ethernet cable, do you still get the dropouts?

Are you using wireless? (and do you have one of those old wireless home phones that interferes? etc.)

 

Your syslog indicates that unRAID is running at GigE speed.

and that your ethernet switch autosenses the speed.

If you're having dropouts of music (tunes typically have very low bandwidth), I'd look at the rest of your network.

For example, if you connect your PC and unRAID via ethernet cable, do you still get the dropouts?

Are you using wireless? (and do you have one of those old wireless home phones that interferes? etc.)

 

I have 4 computers, a smart tv, a ps3 and a media streamer. None of them have any issues if unRaid is out of the equation - everything is hardwired though a gigabit router and a gigabit switch. I only use wireless for the phones and tablets.

 

Additionally, I use a separate workstation (solid state drives only) to download torrents directly onto my unRaid machine (using uTorrent 2.2.1) and I frequently get a "DISK OVERLOAD 100%" error in uTorrent (Yes I get poor performance even when torrents are not running). This also indicates a write performance issue with unRaid (I can get a constant 3MB/sec download speed yet writing to disk can't seem to keep up)

If I were to get a second CPU+motherboard that could hold my drives, could I simply unplug my HDDs, cache drive, and USB key/unRaid OS and plug them into the new machine and boot up to troubleshoot if it has anything to do with the hardware?

If I were to get a second CPU+motherboard that could hold my drives, could I simply unplug my HDDs, cache drive, and USB key/unRaid OS and plug them into the new machine and boot up to troubleshoot if it has anything to do with the hardware?

Yes.

If you can, make sure the BIOS is up to date. (especially if the MB is not new...some older boards have trouble with hard drives >2TB.)

Personally, I always tell everyone to take a 'Screenshot' of their unRAID main web page showing the drive assignments...and store it next to your Passport and Birth Certificate!   unRAID 5 should boot right up and assign all the drives automagically. BUT, if something goes wrong, its very important that you know which drive is your PARITY drive.

I have 4 computers, a smart tv, a ps3 and a media streamer. None of them have any issues if unRaid is out of the equation - everything is hardwired though a gigabit router and a gigabit switch. I only use wireless for the phones and tablets.

 

Additionally, I use a separate workstation (solid state drives only) to download torrents directly onto my unRaid machine (using uTorrent 2.2.1) and I frequently get a "DISK OVERLOAD 100%" error in uTorrent (Yes I get poor performance even when torrents are not running). This also indicates a write performance issue with unRaid (I can get a constant 3MB/sec download speed yet writing to disk can't seem to keep up)

APPEAL for HELP.

I'm stuck for ideas.  The syslog looks fine to me. No plugins. No errors. Network sounds simple...what could be causing network dropouts like this?

 

Smitty2k1: are you sure your utorrents are writing to a user share which is configured to use your cache drive?