unRAID Server Release 5.0-rc10 Available

[Popple2000]

Upgraded to 5.0rc10 from 4.7.

Everything looks great aside from my write speeds now being abysmal(4-5MB/s avg, spikes to 20MB/s), Reads are 80-90MB/s.

Onboard Realtek 8112L

 

Read this thread from this point on:

 

    http://lime-technology.com/forum/index.php?topic=22675.msg220309#msg220309

 

Apparently, some hardware has issues with 32bit OS's and large memory.  (32 bit OS can only directly access 4GB of memory.)

I did, unfortunately it did not fix my issue, but ty

[PeteAron]

OK Here is something I have not seen before.  I am running rc10, and preclearing a 5700 rpm 3 tb drive.  I was trying to copy three 1 gb files to my array, and the share was unresponsive in windows (7).  I cancelled the copy and reinitiated.  Eventually I was able to copy to the array.  Until then the copy dialog box remained open while it calculated time remaining.  From the two failed copy attempts there were a few mb of data written to my drive.  After copying those files successfully I tried again, with the same poor results.  Then I looked at my syslog (enclosed).  Looks like I am running out of memory, but looking at memory info from the system info page:

(from /usr/bin/free)

 

            total      used      free    shared    buffers    cached

Mem:      4132764    1410912    2721852          0    489932    513456

-/+ buffers/cache:    407524    3725240

Swap:            0          0          0

 

It looks like I have plenty. 

 

Here is an example, but there is more in the syslog:

Jan 18 23:09:42 Repository kernel: Pid: 0, comm: swapper/0 Tainted: G          O 3.4.24-unRAID #1 (Errors)

Jan 18 23:09:42 Repository kernel: Call Trace: (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1065e85>] warn_alloc_failed+0xcc/0xdb (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1066769>] __alloc_pages_nodemask+0x4df/0x503 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1085d70>] allocate_slab+0x63/0xc4 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1085df3>] new_slab+0x22/0xeb (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1086358>] T.1697+0x192/0x25c (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12bb598>] ? dev_alloc_skb+0x1c/0x30 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12da2de>] ? ip_local_deliver+0x6d/0x73 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12bb598>] ? dev_alloc_skb+0x1c/0x30 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c108758c>] __kmalloc_track_caller+0xa6/0xf2 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12bb598>] ? dev_alloc_skb+0x1c/0x30 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12ba83a>] __alloc_skb+0x50/0xf5 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12bb598>] dev_alloc_skb+0x1c/0x30 (Errors)

Jan 18 23:09:42 Repository kernel:  [<f84ff8a3>] rtl8168_rx_fill+0x4d/0x106 [r8168] (Errors)

Jan 18 23:09:42 Repository kernel:  [<f84ca11f>] ? rtl8168_get_stats+0xd/0xd [r8168] (Errors)

Jan 18 23:09:42 Repository kernel:  [<f84ffc73>] rtl8168_rx_interrupt+0x317/0x381 [r8168] (Errors)

Jan 18 23:09:42 Repository kernel:  [<f85045f8>] rtl8168_poll+0x2b/0x1cf [r8168] (Errors)

Jan 18 23:09:42 Repository kernel:  [<c11b4cf5>] ? radix_tree_node_rcu_free+0x36/0x38 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12b9dd2>] ? __kfree_skb+0x6a/0x6d (Errors)

Jan 18 23:09:42 Repository kernel:  [<c12c1057>] net_rx_action+0x59/0x12c (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1028619>] __do_softirq+0x6b/0xfa (Errors)

Jan 18 23:09:42 Repository kernel:  [<c10270c1>] ? exit_notify+0x13/0xeb (Errors)

Jan 18 23:09:42 Repository kernel:  [<c10285ae>] ? irq_enter+0x41/0x41 (Errors)

Jan 18 23:09:42 Repository kernel:  <IRQ>  [<c1028462>] ? irq_exit+0x32/0x58

Jan 18 23:09:42 Repository kernel:  [<c1003713>] ? do_IRQ+0x87/0x9b (Errors)

Jan 18 23:09:42 Repository kernel:  [<c133dfa9>] ? common_interrupt+0x29/0x30 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c11fc7f3>] ? acpi_idle_enter_bm+0x276/0x2a9 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c129ce8b>] ? cpuidle_enter+0xe/0x12 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c129d237>] ? cpuidle_idle_call+0x59/0xac (Errors)

Jan 18 23:09:42 Repository kernel:  [<c100897c>] ? cpu_idle+0x41/0x65 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c132c60c>] ? rest_init+0x58/0x5a (Errors)

Jan 18 23:09:42 Repository kernel:  [<c1472816>] ? start_kernel+0x2ab/0x2b1 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c14723e6>] ? kernel_init+0xf8/0xf8 (Errors)

Jan 18 23:09:42 Repository kernel:  [<c14720a8>] ? i386_start_kernel+0xa8/0xaf (Errors)

 

Anyone have any ideas? 

syslog-2013-01-18.txt

[garycase]

kimifelipe  ==>  r.e. your questions about parity check speeds.    If you're getting over 100mb/s you're doing just fine.    Note that a 7200rm 3TB parity drive makes NO difference in your speeds if any other drives are slower (except at the point where all other drives are no longer involved in the check).

 

I wrote the following some time ago that outlines the process.    Note that in your case the 2TB drives you have are "driving" the parity check speed until you're past the 2TB point in the check.  If your only 3TB drive is the parity drive, then speeds after the 2TB point will depend only on it;  but if you have other (slower) 3TB drives installed, they will be the limiting factor for that last TB of checking.

 

As a matter of comparison, I have a system with all 3TB WD Reds ... parity checks take a very consistent 8:05 on this system.

 

A brief note on parity checks ...

 

Remember that all modern drives use zoned sectoring -- so their performance is far better on the outer cylinders than the innermost ones.  So ... regardless of the controller-related performance issues (i.e. PCI controllers vs. PCIe x1 controllers vs. PCIe x4 controllers vs. on-motherboard controllers) ... the performance for a given drive will decrease notably as it reaches the innermost cylinders -- this is most noticeable in the last 15% or so of the drive.   

 

The result of this is that parity checks on a system with mixed drive sizes will slow down much more often than those on systems with all drives the same size.    e.g. if you have 1TB, 1.5TB, and 2TB drives mixed;  the parity check will slow down as it gets to about the 850MB point;  then jump back up in speed at 1TB (as the 1TB drives are now no longer part of the process);  then slow down again as it gets close to the 1.5TB point;  then jump back up until it reaches the last bit of the 2TB drives.

 

Unless you have a slow controller (e.g. a 4-port PCI controller) that can't keep up with the transfer rate of the drives, these drive performance issues are the most significant reason for the parity check rate varying during the process.

[S80_UK]

A brief note on parity checks ...

I could not have explained it better.  Thanks.  I wonder if that explanation should go in the Wiki somewhere.

[PeteAron]

Well I am having trouble copying to my array now.  Windows gives the error "there is a problem accessing \\server\directory....

 

After my problems last night, I allowed the preclear to complete, rebooted the server, and brought the array online.  Opening a server folder is quite slow - it hesitates for 30 to 60 sec, then displays the directory. Attempting to copy a file, though, does not work.  Windows cannot connect to the share. 

 

I ran the permissions script, I might just run it again.

[sdesbure]

Upgraded from rc6a with a realtek NIC (RTL8168C/8111C). It seems that I use r8168. Everything seems to work as fine as before

[Dephcon]

Are there any outstanding NFS issues?  I seem to recall people talking aobut issues with NFS, but I've never used it myself.

 

All my linux htpc friends are telling me i should switch as there's less protocol overhead than smb/cifs and it improves xbmc performance.

[gabbott]

Are there any outstanding NFS issues?  I seem to recall people talking aobut issues with NFS, but I've never used it myself.

 

All my linux htpc friends are telling me i should switch as there's less protocol overhead than smb/cifs and it improves xbmc performance.

 

Unless you are having issues, I wouldn't go through the trouble. A wired connection can run fine with any additional overhead that samba might have.

[IG82]

I have upgraded one of my HP N36L Microservers to RC10.  It was running beta12 prior to this.  This is a spare machine really I am not using properly yet, I still have another running beta12.

 

I ran a parity check prior to and after upgrading.  Both ran clear in terms of parity errors but after upgrading my syslog contains a large number of errors, below is an excerpt.  The parity check takes around 6 hours and these errors only occur for roughly the first half.

 

The only addon I usually use is unmenu and the results are the same when I disable this and run as stock.  The full syslog is also attached.

 

Anyone have any ideas or suggestions?  There are a few other threads but no of these are on RC10.

 

Jan 18 10:18:31 hp2 kernel: INFO: rcu_sched self-detected stall on CPU { 1}  (t=6000 jiffies)
Jan 18 10:18:31 hp2 kernel: Pid: 1210, comm: unraidd Not tainted 3.4.24-unRAID #1 (Errors)
Jan 18 10:18:31 hp2 kernel: Call Trace: (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1057ba2>] print_cpu_stall+0x6d/0xe5 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1057c55>] __rcu_pending+0x3b/0x138 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1057dc8>] rcu_check_callbacks+0x76/0xa1 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c102c96e>] update_process_times+0x2d/0x58 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c104b0fb>] tick_periodic+0x63/0x65 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c104b116>] tick_handle_periodic+0x19/0x6f (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c121d968>] ? add_interrupt_randomness+0x125/0x12d (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c104b467>] tick_do_periodic_broadcast+0x4f/0x88 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c104b4b0>] tick_handle_periodic_broadcast+0x10/0x4b (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1004265>] timer_interrupt+0xd/0x14 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10534a1>] handle_irq_event_percpu+0x2a/0x11a (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1055866>] ? irq_set_chip_and_handler_name+0x27/0x27 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10535b6>] handle_irq_event+0x25/0x3c (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1055866>] ? irq_set_chip_and_handler_name+0x27/0x27 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1055914>] handle_edge_irq+0xae/0xce (Errors)
Jan 18 10:18:31 hp2 kernel:  <IRQ>  [<c10036c3>] ? do_IRQ+0x37/0x9b
Jan 18 10:18:31 hp2 kernel:  [<c1240a62>] ? scsi_finish_command+0x9d/0xa3 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10285ae>] ? irq_enter+0x41/0x41 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c133dfa9>] ? common_interrupt+0x29/0x30 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10285ae>] ? irq_enter+0x41/0x41 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10285ea>] ? __do_softirq+0x3c/0xfa (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10285ae>] ? irq_enter+0x41/0x41 (Errors)
Jan 18 10:18:31 hp2 kernel:  <IRQ>  [<c1028462>] ? irq_exit+0x32/0x58
Jan 18 10:18:31 hp2 kernel:  [<c1003713>] ? do_IRQ+0x87/0x9b (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10401f2>] ? default_wake_function+0xb/0xd (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c133dfa9>] ? common_interrupt+0x29/0x30 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c11b6521>] ? memcmp+0x1a/0x25 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<f851c902>] ? handle_stripe+0xb54/0xdc1 [md_mod] (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1028486>] ? irq_exit+0x56/0x58 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c103d03d>] ? __wake_up+0x3b/0x42 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<f851cbe7>] ? unraidd+0x78/0xbc [md_mod] (Errors)
Jan 18 10:18:31 hp2 kernel:  [<f8519b31>] ? md_thread+0xcc/0xe3 [md_mod] (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10379a1>] ? wake_up_bit+0x5b/0x5b (Errors)
Jan 18 10:18:31 hp2 kernel:  [<f8519a65>] ? import_device+0x166/0x166 [md_mod] (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c10376a8>] ? kthread+0x67/0x6c (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c1037641>] ? kthread_freezable_should_stop+0x49/0x49 (Errors)
Jan 18 10:18:31 hp2 kernel:  [<c133dfb6>] ? kernel_thread_helper+0x6/0xd (Errors)

 

syslog-2013-01-19.zip

[hackztor]

I have been using nfs with unraid to have movies on the Boxee Box and have had 0 issues with nfs. SMB and Boxee Box actually was slower so moved to nfs.

 

Are there any outstanding NFS issues?  I seem to recall people talking aobut issues with NFS, but I've never used it myself.

 

All my linux htpc friends are telling me i should switch as there's less protocol overhead than smb/cifs and it improves xbmc performance.

 

Unless you are having issues, I wouldn't go through the trouble. A wired connection can run fine with any additional overhead that samba might have.

[unevent]

Made the switch to rc10 from rc8a, have Realtek 8111E - no problems.  Ran iperf on rc8a before switch: 112MB/s to/from, ran same test with rc10 and got same values.  Client is Xubuntu 12.10 with Realtek 8111DL.  Have not tested NFS or SMB yet. 

 

Update: Ran some NFS tests with rc10.  I set up a couple 3GB tmpfs - one on unRaid with NFS export and the other on my Xubuntu PC.  Using rsync on Xubuntu and a csv capture with bwm-ng on unraid:

 

File sizes: several different TV shows of 2.2GB each

Ram to Ram tmpfs via NFS: 117MB/s

 

unRaid WD Black to Xubuntu Ram tmpfs via NFS: 103MB/s

(WD Black reported 125MB/s with hdparm -tT)

 

unRaid WD Black to Xubuntu disk via NFS: 65-73MB/s (w/ 1-3 second hiccups and wild fluctuations)

(Disk limited on Xubuntu, hdparm -tT reported 73.76MB/s)

 

Will do SMB next.

 

UPDATE: I ran SMB read/write tests using Windows 7 64bit (Phenom 940, 8GB DDR2, RT8111DL) with 3GB Ram Drive using Dataram RAMDisk Utility.  Used same 3GB tmpfs ram drive on unRaid as used for NFS tests.  unRaid cache SSD is an older Intel X25-V that has not been TRIMed in some time and is getting ate up (hparm -tT shows 116MB/s).  PC Disk is a WD Black SATA3 (for comparison, it's identical twin in unRaid shows 125MB/s with hdparm -tT).  unRaid Parity drive is noted in the sig and the data drive in the array used for writes was the ST1000DM003 at hdparm -tT gave 178MB/s.  Parity gave 145MB/s.  No addons installed (except unmenu).

 

   

   

   

 

To summarize SMB test pics:

PC-Disk to unRaid Cache (no parity) Write: 38.54 MB/s

PC-Disk to Array w/Parity Write: 31.54 MB/s

PC-RAM to unRaid Cache (no parity) Write: 40.37 MB/s

RAM to RAM Read: 116.77 MB/s

RAM to RAM Write: 117.26 MB/s

unRaid Data Disk to PC-RAM Read: 73.70 MB/s

 

---

Unraid 5.0-rc10 - Asus M5A78L-MLX Plus (RT8111E) - AMD Athlon II X3 450 Rana 3.2GHz - 8GB DDR3 - Antec NEO ECO 620W - Antec Three Hundred Case - 1x Rosewill RC-211 (cache) - Parity: 1T Seagate ST1000DM005/HD103SJ - DATA: 3x WD Black 750G - 1x 1T Seagate ST1000DM003 - 1x 500G Seagate ST500DM002, Cache: Intel X25-V SSD 40GB w/8GB swap partition

 

[pras1011]

Which should I run: rc10 or rc10 test?

[Zaxxan]

Which should I run: rc10 or rc10 test?

 

Where can you download rc10 test?

[pras1011]

http://download.lime-technology.com/download/

[Zaxxan]

http://download.lime-technology.com/download/

 

Thanks

[JarDo]

I'm using a modified version of 5.0-rc8a that works with my Realtek card.  I got this version from a member here.  This is a snippet from my syslog taken while running 5.0-rc8a that shows a little info about my card:

 

Jan 19 18:23:58 UNRAID kernel: r8169 Gigabit Ethernet driver 2.3LK-NAPI loaded
Jan 19 18:23:58 UNRAID kernel: r8169 0000:02:07.0: (unregistered net_device): not PCI Express
Jan 19 18:23:58 UNRAID kernel: r8169 0000:02:07.0: eth0: RTL8110s at 0xf8468000, 00:14:c1:4c:38:6f, XID 04000000 IRQ 23
Jan 19 18:23:58 UNRAID kernel: r8169 0000:02:07.0: eth0: jumbo features [frames: 7152 bytes, tx checksumming: ok]

 

Unfortunately, 5.0-rc10 is incompatible with my network card in the same way that the unmodified 5.0-rc8a was.  5.0-rc10 will load, but I can not get to the Web GUI because there is no ethernet.  ifconfig only shows a loopback adapter.

[Joe L.]

Not much will happen if you drop out of the loop after 10 tests/seconds

 

If the UNRAID label does not exist,  the /boot file system will not be mounted, the network will not be configured, and emhttp (the web-interface) will not be started since it is invoked from the config/go script (and the array not started).

 

What do you expect to have happen if the UNRAID label does not exist after 10 seconds. 

(I'm probably missing something that occurs afterwards in the rc.S script, but it is not obvious to me)

 

 

 

[Frank1940]

No, but if the USB system doesn't come up or if it does come but a drive with the lable of 'UNRAID' is not present on it, EXACTLY what you think the computer will be able to do?  With the current situation if the person with problem is looking at the monitor connected to said computer, the message 'waiting for USB subsystem' will be the last line on the screen.  That provides an excellent staring point to find the problem. 

[Joe L.]

Are you sure you don't have access to the system console?

 

I can see the setup of the serial ports comes later in the rc.S script, but I think the system console is already active at that point.  (Watch, I'll be proven wrong  )

[Joe L.]

That's just as helpful to him as 'waiting for USB subsystem' if his USB disk is properly plugged in a USB port.

True, it is a very poorly worded error message.  The USB subsystem could be perfectly fine, but the flash drive label incorrect.

[spants]

 

 

A couple of people have mentioned but it has not been picked up as far as I can see....

 

Shares/nfs keep dropping on 10. I notice it from my xbmc, everything plays fine (a movie). When it has finished and go back the menu to play something else... it doesnt connect. I have to reboot xbmc.

My shares are mainly nfs with some smb.

 

I thought that it was an xbmc problem but just noticed it on a win8 machine that is saving a torrent (legal obviously!) onto an smb directory - error after a while about cant find destination.

 

Will roll back to 8a for more tests.

 

Tony

 

Just to follow up... It wasnt rc10. It was my openelec build...... Now mounting nfs shares in autostart.sh in openelec and it behaves itself.

Not sure why I also got an error on win7, may just have been a one off.

 

[mrjoshzombie]

That looks a lot like a reiserfs file system error leading to a kernel panic.

 

I would perform a reiserfsck on each of your data disks.  You can do this most easily by starting unRAID in maintenance mode (and therefore you do not have to un-mount the disks to perform the tests), you can run them on /dev/md1, /dev/md2, /dev/md3, etc...

 

Joe L.

 

I'd just like to state that this was indeed the issue. I had to run fix-tree on disk2 and all has been well for 3 days now on RC10.

[Mettbrot]

Hey Tom

 

I ran into some problems with permissions to acess files on disks and shares using Windows. Joe L. suggested you may want to take a look at it:

http://lime-technology.com/forum/index.php?topic=25445.0

 

 

I have some questions about permissions:

 

I am running 5.0RC10.

I have changed the name of the server from Tower to Server, dont be confused :-P

I have copied some files to my only disk on the array. I have run the new permissions script to set everything right.

I have created a new user called 'moritz' with no password. I have run the "net use * /delete" command in windows and set up a new entry in the credential manager(\\Server and username moritz)

[*]When I disable user shares, only my disk1 (and the flash drive) is exported via SMB. Now I want to delete a file on the disk, Windows says I need permission from SERVER\nobody. Alright. I set the credentials to \\Server and username to root. Nothing changed. I suppose root is not a valid user for SMB? I don't know. Set it back to moritz.

[*]Enabled user shares and exported the top level folder Movies. Left everything blank, set Security to secure and set user moritz to read/write access. When I now browse to the share via Windows I still cannot delete files. What is even more stange: When I created a new textfile to test I cannot rename it because I need permissions from SERVER\moritz

 

I am realy confused right now 

[limetech]

In rc.S:

# tmm - here's a hack: loop until /dev/disk/by-label/UNRAID is present
# this serves to synchronize this script with USB subsystem
while [ ! -e /dev/disk/by-label/UNRAID ]; do
  echo "waiting for USB subsystem"
  sleep 1
done

 

Do you really need to do that?  What, wait here forever?

 

You could wait a bit -- if you absolutely must -- but then let the machine finish booting up.

 

If shit has happened we need access to the conssole to see/fix what's wrong.

 

(In have a good idea what could be wrong in this particular scenario, but that's beside the point)

 

You don't like my hack? 

 

I guess I was thinking, if the flash doesn't mount there's nothing much to do and if someone actually sees this I would get an email and could investigate what was going on.

 

It's a mean thing to do.  Seriously.

 

Maybe do something like this:

for i in {1..10} ;do
  if [ -e /dev/disk/by-label/UNRAID ] ;then break ;fi
  echo "waiting for USB subsystem..."
  sleep 1
done

.

Please.

 

Ok I can do that, but did you actually see that message repeating on the console, or are you just poking around in the startup files and saw it?  Like I said, it was put in there so that someone would be sure to let me know if it was happening.

[stchas]

I migrated from 4.7 to 5.0-rc10 today and noted that my OCZ Vertex3 SSD reports 0^oC for drive temperature.  It reported temperature correctly under 4.7.

 

Kevin