December 29, 200718 yr I use Unraid 4.2.1 on my server. Essential specs are the same as the MD-1500. WD 1TB drives. The complete network: ------ HTPC - Vista Ultimate 32bit (ASUS P5B Deluxe [onboard Marvell Yukon 88E8056 PCI-E Gigabit], Q6600, 4GB) SERVER - unRAID 4.2.1 (ASUS P5B-VM DO, Celeron 3,46GHz, 2x512MB) ------ Cat6 ------ NETGEAR ProSafe GS105 Gigabit switch ------ Cat6 ------ D-link DIR-655 Gigabit Router ------ Cat6 ------ Workstation (Gigabyte GA-P35-DS4 [onboard Realtek RTL8168/8111 PCI-E Gigabit], Q6600, 4GB) Problem 1: Read performance is really poor using Vista (32bit). Read speed is max 3MB/s. (Write is max 15MB/s) Problem 2: Don't know if this is to expect... But when playing a mp3/avi from the unraid server I get seconds of playback interuption when browsing a share on the unraid server from the vista machine. Also I takes like 20 seconds to enter the video library in Vista Media Center. Problem 3: I can not copy/move from/to the unraid server while streaming video. It causes interuption in playback. Actually a easy task as renaming a folder on the unraid server from the vista or xp machine can take forever if there is a copy or move in progress at the same time. Any ideas?
December 29, 200718 yr I use Unraid 4.2.1 on my server. Essential specs are the same as the MD-1500. WD 1TB drives. The complete network: ------ HTPC - Vista Ultimate 32bit (ASUS P5B Deluxe [onboard Marvell Yukon 88E8056 PCI-E Gigabit], Q6600, 4GB) SERVER - unRAID 4.2.1 (ASUS P5B-VM DO, Celeron 3,46GHz, 2x512MB) ------ Cat6 ------ NETGEAR ProSafe GS105 Gigabit switch ------ Cat6 ------ D-link DIR-655 Gigabit Router ------ Cat6 ------ Workstation (Gigabyte GA-P35-DS4 [onboard Realtek RTL8168/8111 PCI-E Gigabit], Q6600, 4GB) Problem 1: Read performance is really poor using Vista (32bit). Read speed is max 3MB/s. (Write is max 15MB/s) Problem 2: Don't know if this is to expect... But when playing a mp3/avi from the unraid server I get seconds of playback interuption when browsing a share on the unraid server from the vista machine. Also I takes like 20 seconds to enter the video library in Vista Media Center. Problem 3: I can not copy/move from/to the unraid server while streaming video. It causes interuption in playback. Actually a easy task as renaming a folder on the unraid server from the vista or xp machine can take forever if there is a copy or move in progress at the same time. Any ideas? You are not the first to experience poor network performance when using Vista. From what I have read, all three of your issues are the fault of poor networking code in Vista. Good news is, the first service pack for Vista (SP1) is said to cure the networking performance issues. So far a beta and a "RC" (release candidate) version of SP1 are available. install the RC version, or wait till the final release of SP1 occurs and install it then. Joe L.
December 29, 200718 yr Author Yes I'm aware of that. But I thought it should not be so bad... Anyway problem 3 must be something else...as I get the same problem with my XP machine.
December 29, 200718 yr Yes it is SO bad. As for #3 yes probably it is something else. What kind of disks and controllers you use on unRAID?
December 29, 200718 yr Author Yes it is SO bad. As for #3 yes probably it is something else. What kind of disks and controllers you use on unRAID? Onboard controller (ASUS P5B-VM DO) and 2x Promise SATA300 TX4 PCI. 14x Western Digital 1TB (WD10EACS)
December 30, 200718 yr Do you get any errors in the parity check? Could you post your syslog when you're experiencing the issues, that may shed some lights on the problem By the way, how's your speed between XP and Vista?
December 30, 200718 yr It is that bad. Search for our epic thread from a few months back (use my name) if you want all the nasty details . It boils down to Vista not playing well with older versions of Samba...this was even a problem with Vista to XP transfers at one point. The good news is that with RC1 I've gone from ~3M/sec reads to 35M+/sec reads:).
December 30, 200718 yr Bad quote. He said "older" not "old". It is not feasible to incorporate new versions of underlying technology immediately upon their release due to development and testing cycles. Bill
December 30, 200718 yr Tybio "old versions of Samba"? unRAID uses an old version? Why? You have to understand the difference between hacking verses software development. unRAID has a large user base. The first rule of software development is never break what is currently working for your installed user base. That means testing... lots of testing. The latest new version of any component (Samba, TCPIP, NIC driver, etc.) has to have rigorous testing before it is included in the official unRAID release. Note the latest version of unRAID (4.2.1) had only one change from the prior version (4.2) and that was an upgrade of Samba. Most Linux distros don't include the very latest version of many components like Samba either, because they "locked" or "froze" the codebase at a particular level so as to to quality control and things like checking compatibility. You want the latest Samba, then you apply it to your own box, or wait for a new distro to come out. I've played this game with Windows and Window's apps to. I'm both a hacker (not a cracker) and a software developer. A lot of hacking is a "make it work" philosophy that sacrifices quality control and revision testing. It is an important aspect of programming and development, but that paradigm conflicts with much of the paradigm for software development with a large installed user base. I just saw today that a new iPhone update (1.1.3) breaks all third party applications and hacks to unlock the phone to use other networks. Imagine an iPhone update that broke all of Apple's own applications instead of third party ones. THe reason it doesn't is rigorous testing and control on what is and is not included in the package. Tom does a good job with updating unRAID and releasing betas and release candidates, getting feedback, making enhancements and addressing any bugs and incompatibilities that all software upgrades introduce. Look at the frequency of releases and betas between 4.0 and 4.2.1. A 4.3 beta will show up soon.
December 30, 200718 yr BTW, Samba itself is a reverse-engineered application. Microsoft did not release the specifications of the windows file sharing protocol, so the developers of Samba had to completely reverse engineer it. It is a miracle it works at all with Vista since Microsoft made several undocumented changes. In fact, it was less than 2 weeks ago that enough arms were twisted al la the EU antitrust pressures on MS, that MS agreed to provide the protocols under a license, so Samba programmers can actually have the protocols and documentation (over 14,000 pages) to work from. Expect to see more improvements to Samba in the future. See this article: http://www.reuters.com/article/rbssTechMediaTelecomNews/idUSL207546720071220
January 1, 200818 yr Tybio "old versions of Samba"? unRAID uses an old version? Why? You have to understand the difference between hacking verses software development. ... You win for completeness, accuracy, and an overall high degree of helpfulness. I win for brevity - something that those that know me would be surprised to hear. Bill
January 11, 200818 yr Author Another problem...Don't know if this is to be expected. When moving files between disks on the server the remaining time jumps up and down all the time. I store blu-rays and hd dvds and filesize is about 30GB most of the time. To move a 30GB file from one disk to another on the server takes about 30 minutes. Can't be right? If using explorer to browse the files everything get slugish. If you try to drag and drop a file from one disk to another it takes like 5 seconds before the mousepointer moves to the other window (with the other disk) and 10 seconds after drop until it start to move the file. Then remaining time is pending between 20 and 40 minutes. Usually it takes about 30 minutes to complete the move. Tested from 2 diffrent computers both running clean installations of Windows XP SP2. When moving a file from one disk to another my XP boxs network performance jumps up and down between 0% and 30% (gigabit). Does this mean that the file is actually moved through my XP machine and not internally on the unraid server? Sometimes I get an error halway through a copy. Translated from swedish to something like "The provided nework path does no longer exist".
January 11, 200818 yr As Darts said, we need to see your syslog. Also, we need to know how you are doing the moves and copies. If they are from the Windows machine, then Windows is doing the copy/move from one drive (local or networked) to the other drive (local or remote), so it is crossing the wire twice in your case. If you are doing them using Linux commands from the console or a Telnet box, then they should be much faster. Thirty minutes for 30GB could be about right, from Windows.
January 11, 200818 yr Good to see attachments working again! Your section of the syslog is only the last third of it, but enough to say there does not appear to be anything wrong. I recommend copying and attaching the entire syslog. Also, try doing a big copy, and then capturing the syslog afterward, to see if errors appear in it.
January 11, 200818 yr Author Good to see attachments working again! Your section of the syslog is only the last third of it, but enough to say there does not appear to be anything wrong. I recommend copying and attaching the entire syslog. Also, try doing a big copy, and then capturing the syslog afterward, to see if errors appear in it. Can someone please tell me the right way to copy the whole syslog.
January 11, 200818 yr Author There you go Thanks! For some reason the syslog can't be copied to the flash share. But it did work with disk1. Here it is... Hope you can find something that is wrong. Been trying to copy a 40GB file from my XP machine to the unraid server 6 times now this evening. But it fails after about 30 minutes...saying that the network path is incorrect.
January 12, 200818 yr I think this may be the problem : Jan 11 13:29:58 UNRAID kernel: [ 8440.479086] ata6.00: exception Emask 0x2 SAct 0x1cfa SErr 0x0 action 0x2 frozen Jan 11 13:29:58 UNRAID kernel: [ 8440.479092] ata6.00: (spurious completions during NCQ issue=0x0 SAct=0x1cfa FIS=004040a1:00000001) Jan 11 13:29:58 UNRAID kernel: [ 8440.479099] ata6.00: cmd 61/08:08:af:ed:f1/00:00:43:00:00/40 tag 1 cdb 0x0 data 4096 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479100] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479106] ata6.00: cmd 61/08:18:57:f7:f1/00:00:43:00:00/40 tag 3 cdb 0x0 data 4096 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479107] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479113] ata6.00: cmd 61/d0:20:87:f8:f1/00:00:43:00:00/40 tag 4 cdb 0x0 data 106496 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479114] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479120] ata6.00: cmd 61/18:28:b7:ed:f1/00:00:43:00:00/40 tag 5 cdb 0x0 data 12288 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479121] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479127] ata6.00: cmd 61/28:30:87:f7:f1/00:00:43:00:00/40 tag 6 cdb 0x0 data 20480 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479128] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479133] ata6.00: cmd 61/b0:38:af:f7:f1/00:00:43:00:00/40 tag 7 cdb 0x0 data 90112 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479135] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479140] ata6.00: cmd 61/28:50:5f:f8:f1/00:00:43:00:00/40 tag 10 cdb 0x0 data 20480 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479142] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479147] ata6.00: cmd 61/08:58:4f:f7:f1/00:00:43:00:00/40 tag 11 cdb 0x0 data 4096 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479149] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.479154] ata6.00: cmd 61/28:60:5f:f7:f1/00:00:43:00:00/40 tag 12 cdb 0x0 data 20480 out Jan 11 13:29:58 UNRAID kernel: [ 8440.479156] res 40/00:44:cf:f9:f1/00:00:43:00:00/40 Emask 0x2 (HSM violation) Jan 11 13:29:58 UNRAID kernel: [ 8440.808475] ata6: soft resetting port Googleing it indicates it might be an incompatibility between NCQ and Kernel 2.6.19, unRAID 4.2.1 uses 2.6.22.5 but I didn't see the issue with that kernel. It also might be an issue with your controller but that's as far as I can see for the moment... ... Hmmm, This looks interesting... Looks like they made a kernel patch for this issue!
January 12, 200818 yr Your syslog looks OK, except for one drive exception and 2 network dropouts. The exception involves a series of HSM violations (discussed above and elsewhere by Tom), and seems to be an anomaly here, not particularly serious, and didn't cause a delay. It did cause a reset of the port, and that could potentially be a problem, a minor hiccup. The 2 network connections lost were for about 5 seconds and about 9 seconds, which could definitely cause the network errors you are seeing. There were NO errors on your unRAID box, other than reporting the link was down, so that leads me to suspect you have a problem elsewhere in your network, intermittent loss of networking. Possibly a router or switch problem? The speeds and start/stop behavior you are seeing seem normal and reasonable, as you are somewhat PCI bound. One disk is read, and the data is sent across the network to the XP machine, returns from the XP machine, then the destination drive is read then written, and the parity drive is read and written. You might want to study the Linux versions of copying and moving, the cp and mv commands, as they would eliminate the round trip to the XP station and the potential network dropouts. You cannot eliminate the PCI and unRAID writing bottleneck though.
January 12, 200818 yr Author Thanks for your replies guys! I'm really thankful for all help. In the link from Dart... He says the system gets unresponsive. Thats about what I experience. I'm using the same motherboard and controllers as in Tom's MD-1500. I can move small files to the server without problems. So something tells me that there is a connection loss at some intervall so the bigger files does not have time to finish. Hmm...I tried to remove the switch (look at first post for setup) but no luck. There is no settings in unraid to choose link speed? I have router, and XP machine set to auto. NIC settings on XP machine (Realtek RTL8168/8111 PCI-E gigabit integrated): Flow Control: Disabled Jumbo Frame: Disabled Link Speed/Duplex Mode: Auto Negotiation Offload Checksum: Tx/Rx Checksum Offload TCP_LargeSend: Enable No idea if this can affect performance...don't know that much about networking...
Archived
This topic is now archived and is closed to further replies.