August 30, 200916 yr Anyone tried this puppy? I got one of these a month ago, and it's been runnung pretty stable so far. (Bubbaraid+unmenu) There are some things that bug the hell out of me though: -- 1. The disks run slow: My new super-duper WD-RE4 SATA-II disk behaves like ATA/100. And my new 2.5 inch WD-Scorpio Blue runs as ATA/33. -- 2. Whenever two or more disks are accessed at the same time, the system gets really, REALLY slow. If you look at htop at times like this you see a whole bunch of processes stuck in device-wait state for a long time. -- 3. BIOS claims (and also the spec-sheets claim) that it can run memory at DDR200, DDR266, and DDR333. I've not been able to get DDR333 though. I tried 6 different sticks of memory so far, some have been proven to run DDR333 on other computers, and the highest this box goes is DDR266. Despite the above three annoyances, it is still a pretty good deal for 300 bucks! It eats about 30 Watts at the plug when in heavy use, and for someone who only uses this to watch DVD movies off of it, it does a great job! Still I'm wondering if something can be chnged in the unraid settings to fix the disk performance. I am not too optimistic though: I suspect that this box has the ICH4 chipset, and (I hope my guess is wrong) the ICH4 chipset doesn't have a good SATA controller on board. For those of you who understand Linux, I attach my syslog here. Maybe somebody will see something wrong in it and come up with a suggestion. Thanks! Purko This snippet complains about the 80wire PATA cable and drops the speed to UDMA33. Check the cable on this one. Aug 28 12:58:35 Tower kernel: hda: WDC WD3200BEVE-00A0HT0, ATA DISK drive Aug 28 12:58:35 Tower kernel: hda: host max PIO4 wanted PIO255(auto-tune) selected PIO4 Aug 28 12:58:35 Tower kernel: hda: host side 80-wire cable detection failed, limiting max speed to UDMA33 Aug 28 12:58:35 Tower kernel: hda: UDMA/33 mode selected Aug 28 12:58:35 Tower kernel: ide0 at 0x170-0x177,0x376 on irq 15 Aug 28 12:58:35 Tower kernel: hda: max request size: 512KiB Aug 28 12:58:35 Tower kernel: hda: 625142448 sectors (320072 MB) w/8192KiB Cache, CHS=38913/255/63 Aug 28 12:58:35 Tower kernel: hda: cache flushes supported Aug 28 12:58:35 Tower kernel: hdhttp://lime-technology.com/forum/index.php?action=post;quote=37662;topic=3453.0;sesc=2f938e89279bf70d6fab0ff847ca803fa: hda1 .... Aug 28 12:58:35 Tower kernel: ata5: SATA link up 3.0 Gbps (SStatus 123 SControl 300) Aug 28 12:58:35 Tower kernel: ata5.00: HPA detected: current 3907029168, native 18446744073321613488 Aug 28 12:58:35 Tower kernel: ata5.00: ATA-8: WDC WD2002FYPS-01U1B0, 04.05G04, max UDMA/133 Aug 28 12:58:35 Tower kernel: ata5.00: 3907029168 sectors, multi 0: LBA48 NCQ (depth 31/32) Aug 28 12:58:35 Tower kernel: ata5.00: max_sectors limited to 256 for NCQ Aug 28 12:58:35 Tower kernel: ata5.00: max_sectors limited to 256 for NCQ Aug 28 12:58:35 Tower kernel: ata5.00: configured for UDMA/133 .... For the other drives, they seem to be initializing correctly. Only thing I noticed is they have NCQ enabled. I though there was a performance issue on the latest unRAID and it was suggested to set NCQ to 1 or 0. I cannot remember. When you do a parity check, what is the speed? I noticed there were two pre-clears started up close together. that would bog the system, but it should not be that bad. This is using the sata_nv driver for the SATA disks. Also, it supports port multipliers. In theory you could have 20 disks on this. (probably not practical, but it could work). Where did you get this for $300? It is a sweet deal.
September 1, 200916 yr This snippet complains about the 80wire PATA cable and drops the speed to UDMA33. Check the cable on this one. That's the tricky part! This disk can't have 80-pin cable (or 40-pin for that matter). It's a 2.5" mini-IDE, and they have 44-pin cables only. As seen in: http://www.kernel.org/doc/Documentation/kernel-parameters.txt I tried adding libata.force=80c in syslinux.cfg, but the system replies with "unknown command: libata.force=80c, ignoring." Any ideas how I can make the system ignore the cable test? For the other drives, they seem to be initializing correctly. Only thing I noticed is they have NCQ enabled. I though there was a performance issue on the latest unRAID and it was suggested to set NCQ to 1 or 0. I cannot remember. I wouldn't know how to disable/enable NCQ. I tried adding libata.force=noncq in syslinux.cfg, but this doesn't work for some reason. I believe the latest 4.5.x betas have a switch to disable it, but I am running 4.4.2 because that's as high as BubbaRaid would go. Where did you get this for $300? It is a sweet deal. I followed the e-bay link in the first page of this thread. It is indeed a sweet deal! Currently I have 12TB+ storage in there! -- 4x 2TB WD Green -- 1x 2TB WD RE-4 (for parity) -- 1x 320GB WD corpio 2.5" (not in the protected array -- for scratch, and for rTorrent) -- 1x 2TB WD external USB -- 3x still unused eSATA ports for further expansion. Can't beat that! Purko
September 2, 200916 yr This snippet complains about the 80wire PATA cable and drops the speed to UDMA33. Check the cable on this one. That's the tricky part! This disk can't have 80-pin cable (or 40-pin for that matter). It's a 2.5" mini-IDE, and they have 44-pin cables only. As seen in: http://www.kernel.org/doc/Documentation/kernel-parameters.txt I tried adding libata.force=80c in syslinux.cfg, but the system replies with "unknown command: libata.force=80c, ignoring." Any ideas how I can make the system ignore the cable test? I did a little research, and there are a few combinations of kernel and drive that appear to have a buggy cable detection. A different Linux kernel may work better for you. Also, just to verify, did you add libata.force=80c to the line that begins with append? From what I read, short40c may be better for you than 80c, but that's a moot point if the libata.force patch module is not included in this release. Only thing I noticed is they have NCQ enabled. I though there was a performance issue on the latest unRAID and it was suggested to set NCQ to 1 or 0. I cannot remember. I wouldn't know how to disable/enable NCQ. I tried adding libata.force=noncq in syslinux.cfg, but this doesn't work for some reason. I believe the latest 4.5.x betas have a switch to disable it, but I am running 4.4.2 because that's as high as BubbaRaid would go. There is a little tool, found here, that will let you play with toggling NCQ on and off.
September 8, 200916 yr -- 2. Whenever two or more disks are accessed at the same time, the system gets really, REALLY slow. If you look at htop at times like this you see a whole bunch of processes stuck in device-wait state for a long time. I am verry happy to report that the I/O problem was solved! Problem Description: I/O operations on large files tended to produce extremely high iowait times and poor system I/O performance. Copying lots of gigabytes of files rendered my unRAID almost unuseable: Even starting a new shell was taking a long time, and sometimes that even failed with a message about wrong user/pass. The samba client also became unusable at such time. CPU hit 100%. Solution: It turned out that the default I/O scheduler in unRAID is noop. Changing that to any of the other available schedulers (anticipatory, deadline, or cfq) completely cured the problem, and brought a drastic improvement to the system's multitasking performance. A brief description of the IO schedulers can be found here: http://www.linuxhowtos.org/System/iosched.htm At the moment, I have all my hard disks switched to cfq. I am testing the server with as much load as I can put on it, and I get no deadlocks whatsoever. I have started: -- Parity-Sync on the array; -- Two concurrent preclear scripts on two new disks; -- A mc copy of 120GB folder to my scrstch disk (outside the array); -- A samba copy of a 50GB folder from a windows pc; -- From another win PC watching a DVD off the array; -- A few telnet sessions, one of them watching htop. ....and with all of the above running at the same time, this tiny little mobile CPU never hit 80%! Beautiful! Yours, Purko
January 11, 201016 yr -- 2. Whenever two or more disks are accessed at the same time, the system gets really, REALLY slow. If you look at htop at times like this you see a whole bunch of processes stuck in device-wait state for a long time. I am verry happy to report that the I/O problem was solved! Problem Description: I/O operations on large files tended to produce extremely high iowait times and poor system I/O performance. Copying lots of gigabytes of files rendered my unRAID almost unuseable: Even starting a new shell was taking a long time, and sometimes that even failed with a message about wrong user/pass. The samba client also became unusable at such time. CPU hit 100%. Solution: It turned out that the default I/O scheduler in unRAID is noop. Changing that to any of the other available schedulers (anticipatory, deadline, or cfq) completely cured the problem, and brought a drastic improvement to the system's multitasking performance. A brief description of the IO schedulers can be found here: http://www.linuxhowtos.org/System/iosched.htm At the moment, I have all my hard disks switched to cfq. I am testing the server with as much load as I can put on it, and I get no deadlocks whatsoever. I have started: -- Parity-Sync on the array; -- Two concurrent preclear scripts on two new disks; -- A mc copy of 120GB folder to my scrstch disk (outside the array); -- A samba copy of a 50GB folder from a windows pc; -- From another win PC watching a DVD off the array; -- A few telnet sessions, one of them watching htop. ....and with all of the above running at the same time, this tiny little mobile CPU never hit 80%! Beautiful! Yours, Purko Was it difficult to change the scheduler? How much UNIX knowledge is required?
January 11, 201016 yr A brief description of the IO schedulers can be found here: http://www.linuxhowtos.org/System/iosched.htm Was it difficult to change the scheduler? How much UNIX knowledge is required? The link he gives in his post expains how to change it via the command line. It is pretty easy to do, and all you would need to do is add that to your go script to make sure it is changed on every reboot (assuming that is needed).
January 11, 201016 yr Thanks prostuff1! I should have seen that for myself. A brief description of the IO schedulers can be found here: http://www.linuxhowtos.org/System/iosched.htm Was it difficult to change the scheduler? How much UNIX knowledge is required? The link he gives in his post expains how to change it via the command line. It is pretty easy to do, and all you would need to do is add that to your go script to make sure it is changed on every reboot (assuming that is needed).
April 16, 201016 yr Hi all, I've been lurking here for a month or so, and this is my first post. Nice to meet you. I just pulled the trigger on this: http://cgi.ebay.com/Norco-DS-520G_W0QQitemZ320515834413QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item4aa03b822d#ht_500wt_1176 Any other 520 owners out there have any advice on what I should do special for this box? Bios version, bios settings changes, that sort of stuff? Thanks!
April 16, 201016 yr Any other 520 owners out there have any advice on what I should do special for this box? Bios version, bios settings changes, that sort of stuff? I have the NS-520, not the DS. But from the specs you listed there, they seem to be exactly the same. I am very happy with mine! I have it populated with 6 disks, and I've also connected another 5-disk eSATA enclosure to it. Here are a few quirks I've discovered so far: 1. Boot from USB: I wasn't able to make mine boot from the USB flash key, until I went in BIOS and disabled IDE. When the server booted, it was still able to recognize and use the 2.5" EIDE disk that I have in there, regardles of BIOS. 2. Initilly the server was selecting UDMA/33 mode for the 2.5" EIDE disk, because it couldn't detect a 80-pin IDE cable. (this cable is a short 44-pin, including the power pins). My workaround is too boot with the "ide_core.ignore_cable=0" boot option. So now when the server boots, it first tries UDMA/133 mode for it, which of course fails, and then they settrle on UDMA/66 mode. From that point on the disk works perfectly well. I just ignore those few "error" messages during boot. 3. ACPI: When the server first booted, it wrote a line in syslog saying that this motherboard/BIOS doesn't support ACPI. So I tried booting the server with the "acpi=force" boot option, and ACPI worked perfectly well, without any problems. Then, a few months later I read somewhere on the web that using "acpi=force" is not a good idea with ICH4 chipsets, so I stopped using that. 4. The I/O scheduler: I've found that CFQ works better than NOOP on this box, so I boot it with the "elevator=cfq" option. That, and some other tweaks that have shown to improve performance on this box can be found in my 'go' script posted here: http://lime-technology.com/forum/index.php?topic=5976.msg56946#msg56946 You made a good purchase! I am eagerly waiting to hear how things go with your server. Purko
April 16, 201016 yr Hi all, I've been lurking here for a month or so, and this is my first post. Nice to meet you. I just pulled the trigger on this: http://cgi.ebay.com/Norco-DS-520G_W0QQitemZ320515834413QQcmdZViewItemQQptZLH_DefaultDomain_0?hash=item4aa03b822d#ht_500wt_1176 Any other 520 owners out there have any advice on what I should do special for this box? Bios version, bios settings changes, that sort of stuff? Thanks! Great Price.. Good thing I did not see it. I would have scoffed it up!!! LOL!!! I like this lil machine. I also have an external 5 drive unit for it. With the 5 drive units, this machine is capable of 20 drives, although I bet it will be a bit slow and require a round robin approach to assigning the drives.
April 16, 201016 yr Purko, does your unit power off when you do a powerdown or shutdown. Mine seems to just sit there with "powering off" but it never turns off.
April 16, 201016 yr Purko, does your unit power off when you do a powerdown or shutdown. Mine seems to just sit there with "powering off" but it never turns off. It powers down through software command if I've booted it with the "acpi=force" option. (my previous post) If it's booted without ACPI, the machine just halts at the end and you need to push the button.
April 16, 201016 yr Purko, does your unit power off when you do a powerdown or shutdown. Mine seems to just sit there with "powering off" but it never turns off. It powers down through software command if I've booted it with the "acpi=force" option. (my previous post) If it's booted without ACPI, the machine just halts at the end and you need to push the button. OK, That's what I thought, But I just wanted to check. I've been wanting to use my DS-520G as a pure backup solution where it powers up, runs the rsyncs, then powers off.. But it never turns itself off.
April 16, 201016 yr I like this lil machine. I also have an external 5 drive unit for it. With the 5 drive units, this machine is capable of 20 drives... 22 when you count the IDE slots.
April 16, 201016 yr I like this lil machine. I also have an external 5 drive unit for it. With the 5 drive units, this machine is capable of 20 drives... 22 when you count the IDE slots. I don't count them. although I do have a 2.5" IDE drive, the size of them makes me only consider it for a cache drive. The other IDE (CF slot) I have enabled, but I'm not using at the moment. I had thought about changing the boot logic to boot from IDE CF until I started working with grub4dos and I was able to boot from any drive very easily.
April 28, 201016 yr Ok guys! All my stuff is here! I have received: (1) 2tb wd black for parity (1) 2tb wd green usb key from Tom. (Thanks!) and my Norco 520 from ebay. I have another 10 drives in 2 multiplier chassis' that have data on them, so those will have to be copied over 1 at a time. I'm hooking up a monitor and getting ready to power up. I guess I need to hit the bios first. If anyone with a Norco 520 wants to help me on this thread tonight, I'm in it for the long haul to get this sucker going! Wish me luck.
May 1, 201016 yr The Norco is up and functioning. I've applied all tweaks listed in this thread. I'm rsync'ing some files over as I type this and I am getting around 5MB/s. My network is all cat6 with all gigE ports and switches, _except_ right here where the Norco is on my bench. Do you guys believe 5MB/s is sane for 100mbit lan? What do you guess my write speed would be if I go snatch up another gigE switch? Oh yeah unMenu rocks _real_ hard.
May 2, 201016 yr I'm back and a gigE switch is in place. Now I'm averaging 9 or 10 MB/s. What do y'all think I should be getting? What should I troubleshoot? Thanks
May 2, 201016 yr I'm back and a gigE switch is in place. Now I'm averaging 9 or 10 MB/s. What do y'all think I should be getting? What should I troubleshoot? Thanks You should troubleshoot either your LAN or your Server. Most people are reporting write speeds from 30 to 40MB/s if they are using Gigabyte LAN, high rotational speed SATA disks (7200 RPM) and the 4.5.3 release of unRAID. It sounds as if you are still connecting at 100 MB/s somewhere between the PC and the server. If you type ethtool eth0 what speed does it report? Joe L.
May 2, 201016 yr ethtool output: root@unraid:~# ethtool eth0 Settings for eth0: Supported ports: [ TP ] Supported link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Supports auto-negotiation: Yes Advertised link modes: 10baseT/Half 10baseT/Full 100baseT/Half 100baseT/Full 1000baseT/Full Advertised auto-negotiation: Yes Speed: 1000Mb/s Duplex: Full Port: Twisted Pair PHYAD: 0 Transceiver: internal Auto-negotiation: on Supports Wake-on: umbg Wake-on: d Current message level: 0x00000007 (7) Link detected: yes root@unraid:~# network topology: +-------------+ |10/100 router| +------+------+ | +---------+--------+ |10/100/1000 switch| +---------+--------+ | +---------+--------+ |10/100/1000 switch| +-----+-------+----+ | | | | +---------------+--+ +--+-------------------+ |box with the goods| |the unraid destination| +------------------+ +----------------------+ ping: loki:~ xaos01$ ping 192.168.1.119 PING 192.168.1.119 (192.168.1.119): 56 data bytes 64 bytes from 192.168.1.119: icmp_seq=0 ttl=64 time=0.415 ms 64 bytes from 192.168.1.119: icmp_seq=1 ttl=64 time=0.218 ms 64 bytes from 192.168.1.119: icmp_seq=2 ttl=64 time=0.364 ms 64 bytes from 192.168.1.119: icmp_seq=3 ttl=64 time=0.357 ms ^C --- 192.168.1.119 ping statistics --- 4 packets transmitted, 4 packets received, 0% packet loss round-trip min/avg/max/stddev = 0.218/0.339/0.415/0.073 ms rsync: loki:~ xaos01$ rsync -avP --exclude '.DS*' --exclude '._*' /Volumes/drive2/backup/ /Volumes/disk1/movies/ building file list ... 89 files to consider ./ 30 Days of Night (2007)/ 30 Days of Night (2007)/30 Days of Night (2007).mkv 7349317635 100% 7.77MB/s 0:15:01 (xfer#1, to-check=84/89) 30 Days of Night (2007)/30 Days of Night (2007).nfo 2272 100% 20.74kB/s 0:00:00 (xfer#2, to-check=83/89) 30 Days of Night (2007)/30 Days of Night (2007).tbn 848630 100% 1.89MB/s 0:00:00 (xfer#3, to-check=82/89) Bubba Ho-tep (2002)/ Bubba Ho-tep (2002)/Bubba Ho-tep (2002)-fanart.jpg 243886 100% 487.06kB/s 0:00:00 (xfer#4, to-check=80/89) Bubba Ho-tep (2002)/Bubba Ho-tep (2002).mkv 4693375536 100% 7.19MB/s 0:10:22 (xfer#5, to-check=79/89) Bubba Ho-tep (2002)/Bubba Ho-tep (2002).nfo 2624 100% 4.79kB/s 0:00:00 (xfer#6, to-check=78/89) Bubba Ho-tep (2002)/Bubba Ho-tep (2002).tbn 452978 100% 563.52kB/s 0:00:00 (xfer#7, to-check=77/89) Cast Away (2000)/ Cast Away (2000)/Cast Away (2000)-fanart.jpg 550129 100% 558.46kB/s 0:00:00 (xfer#8, to-check=75/89) Cast Away (2000)/Cast Away (2000).mkv 10370187264 73% 7.23MB/s 0:08:23 Disks: Source == Samsung 1TB HD103UJ unraid destination == 2TB Green WDC_WD20EARS pins jumpered. unraid parity == 2TB Black WDC_WD2001FASS Thanks for any and all help!
May 12, 201016 yr The plot thickens... Ok so I'm doing testing, engaging the network engineers where I work, and they tell me to initiate an rsync from within my Norco from disk1 to disk2. Avoiding the network as well as SMB. Same wall. ~10MB/s. So now I know it's not my network and not SMB. Is my Norco busted? My drives? Sure sounds like there's a misconfiguration somewhere. This was between 2 WD 2TB EARS Green drives (properly jumpered) with a WD 2TB Caviar Black parity. Anyone have any ideas what I should check? Thanks
May 12, 201016 yr Would be nice if another NS-520 owner could comment on their performance. It's possible the controller just sucks, and is not capable of doing better. I bought a WD MyBook World Edition maybe 8 months ago, and the NAS chip in it was only capable of delivering about 1.5MB/s write speeds. I ended up returning it based on that. I suppose it's possible you're in a similar situation. I'd love to hear from another owner, as I was just thinking to pick one of these up for offline backups. Power up, backup, power down.
Archived
This topic is now archived and is closed to further replies.