I've finally got all my drives precleared and ready to go. I've begun the transfer of files from my old server to the unRAID, but its painfully slow. 4-7 MB/s. When I was copying from one of the desktops to the old server I would get about 70-75MB/s. My setup is as follows.

 

• unRAID connect w/GB to switch A
  
• Switch A connects to switch B in another part of the house
  
• Old server is connected to switch B
  

 

So basically there are two GB switches between the old server and the unRAID.  I noticed that the speed has not changed since I added the cache drive. I know its using it cause I can see the data on there as its transferring. Here is the ethtool and ifconfig output.

 

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 pause frame use: No

        Advertised auto-negotiation: Yes

        Speed: 1000Mb/s

        Duplex: Full

        Port: Twisted Pair

        PHYAD: 1

        Transceiver: internal

        Auto-negotiation: on

        MDI-X: off

        Supports Wake-on: pumbg

        Wake-on: g

        Current message level: 0x00000007 (7)

        Link detected: yes

 

 

eth0      Link encap:Ethernet  HWaddr 00:1e:4f:a1:34:50

          inet addr:10.10.10.156  Bcast:10.10.10.255  Mask:255.255.255.0

          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1

          RX packets:5178762 errors:0 dropped:185 overruns:0 frame:0

          TX packets:3051956 errors:0 dropped:0 overruns:0 carrier:0

          collisions:0 txqueuelen:1000

          RX bytes:7221645572 (6.7 GiB)  TX bytes:213057378 (203.1 MiB)

          Interrupt:21 Memory:fe9e0000-fea00000

 

Is there something obvious I'm missing? I understand unRAID isn't really meant for performance, which is fine, but I'm not even getting 10MB/s.

 

Thanks

 

P.S. Bonus question. My cache drive is 750GB, what happens if I'm transferring over a TB and the cache fills up? Will it automatically dump that the array and carry on?

 

If both switches are gigabit and the inter-switch connection is negotiated as gigabit, then it sounds like the old server only has a 100Mbit NIC, or auto-negotiated at 100Mbit (assuming a gigabit NIC).

 

I believe if the cache drive fills up, the transfer is halted as there is no more free space.  It does not move the existing cache to the array and then continue.

 

EDIT:  The desktop that you say is getting 75MB/s, is that on switch A or switch B?  If it's on switch B, that rules out any issues with the inter-switch connection.

I believe if the cache drive fills up, the transfer is halted as there is no more free space.  It does not move the existing cache to the array and then continue.

Not true, unless things aren't configured correctly. The next file that would overrun the free space on the cache will be written directly to the user share, following whatever disk allocation, split level, and free space rules are in place. Now, if you are writing directly to /mnt/cache, then yes, it will fill up and abort the transfer of the next file that won't fit. You can always run the mover manually at any time though.

I doubled checked the server nic and it is set at 1GB. The desktop comp is actually on a GB router. I suppose I could try going directly from the same switch the server is on the the unRAID, it would just require some work. Seems strange that its that slow though regardless.

I believe if the cache drive fills up, the transfer is halted as there is no more free space.  It does not move the existing cache to the array and then continue.

Not true, unless things aren't configured correctly. The next file that would overrun the free space on the cache will be written directly to the user share, following whatever disk allocation, split level, and free space rules are in place. Now, if you are writing directly to /mnt/cache, then yes, it will fill up and abort the transfer of the next file that won't fit. You can always run the mover manually at any time though.

 

I didn't know that, thanks for setting me straight.

I doubled checked the server nic and it is set at 1GB. The desktop comp is actually on a GB router. I suppose I could try going directly from the same switch the server is on the the unRAID, it would just require some work. Seems strange that its that slow though regardless.

 

Is it at 1Gb, full duplex?  Have you done a transfer to unRAID from any other machine located on the same switch as the old server?  If your desktop machine transferred to unRAID at 75MB/s, that definitely rules out unRAID as the issue.  The issue lies in either the old server's switch connection, or the inter-switch connection.  A successful high-speed transfer from any other machine with a 1Gb NIC on the same switch as the old server (switch B) would rule out the inter-switch connection.

 

Actually it's not that strange that it is that slow if either the old server or the inter-switch connection is at 100 Mbit, especially if it's at 100Mbit half-duplex.  100Mb/s (12.5MB/s) is theoretical, not real-world.  You'll typically see a bit less than that, but 10MB/s should easily be doable at 100Mb/s full-duplex.

 

EDIT:  Also, what cables are you using?  CAT5, 5e, 6, 6e?  Purchased, or did you crimp them yourself?  CAT5 won't cut it for gigabit, you need CAT5e or above.  Every cable between the old server and unRAID must be CAT5e or better, including the inter-switch connection.

 

Is it at 1Gb, full duplex?  Have you done a transfer to unRAID from any other machine located on the same switch as the old server?  If your desktop machine transferred to unRAID at 75MB/s, that definitely rules out unRAID as the issue.

 

It is set to 1Gb full duplex. The transfer speeds I got at 75MB/s were between another desktop computer and my old WHS server. So far I haven't been able to get anything past 7MB/s on unRAID. I'll check and make sure all cables connected to my system are CAT5e or above.

Ah, my mistake, I misread your OP.  Have you tried a transfer to unRAID from another machine on the same switch?  What are your system specs, specifically mobo and RAM?  There is a known issue with certain motherboards and greater than 4GB of RAM that sometimes leads to slow writes.

The system is a Dell Optiplex 755 with a Quad Core and 8GB of RAM.

 

On that note, I'm a tool. After looking at the cabling I realized I plugged the unRAID into a 100MB port, rather than one of the GB ports. I've now taking the cable I plugged into the GB port and tested it on my laptop to make sure it is in fact GB, it is. I checked the unRAID box now and its showing as being connected at 100MB Full Duplex. It says that 1000MB Full is a supported option so I tried this.

 

ethtool -s eth0 duplex full speed 1000 autoneg off

 

I check it again after a couple mins and its now saying something about Speed and Duplex being unknown and it won't give me an IP. AHHH!

 

Any ideas. Thanks for the help.

Well, I broke the first rule in IT, test the cable. The cable I'm using is CAT6 from a few I bought a few weeks back. Installed it and everything was running fine. Added the unRAID and it was not running at GB speeds. Didn't think much about the cable because I plugged it into a laptop and it came up GB. After trying everything else I could think of I tried another cable and.....BOOM! Works. SOB! Strange that it worked for the laptop, but not unRAID. Mayb its more sensitive or something? 

 

Anyway, thanks for the help.

 

P.S. Still only getting about 45MB/s, but I may be able to tune that up a bit.

Glad you got it sorted out. 40 - 45 MB/s is about all you'll get (even with fast spinners) on writes to the parity protected array, due to the overhead of parity calculation. That said, if you are writing to the cache drive instead of the protected array, you should be able to hit the drive at its max write speed or saturate the GB connection, whichever is lower.

Still only getting about 45MB/s, but I may be able to tune that up a bit.

 

I wish I could get that.  I max out at about 35 MB/s.

45MB would be VERY good IF you were writing to the protected array ... but it's not so good when writing to the cache drive unless the cache drive is fairly full (i.e. you're writing to the inner cylinders).  But if this is an older drive (especially if it's IDE) that may be a reasonable speed.    Also, are you writing with Explorer or using another copy utility (e.g. TeraCopy)?

 

As for the speed issue => Low transfer speeds when you know all the NICs, switches, and routers are gigabit  are ALMOST (never say never) always due to one of two things:

 

(a)  A defective cable or connector.  When it's still working, but just very slow, it's almost always an issue with one of the pairs on the cable or a defective ground.    This can cause a very high number of packet retries, thus resulting in a very slow transfer speed.    This can be caused by an actual break in the cable (usually caused by bending it too much);  or can be a simple case where the cable isn't fully seated in the connector on one or more pins.

 

or

 

(b)  A defective port on your switch or router.    I've seen far too many of these -- where one port goes bad and is suddenly transferring at very low speeds, even though the other ports on the same switch are fine.  I have a couple of "4 port switches" in my junk collection (5 port GB switches with one bad port).

 

You also need to watch cheap consumer switches.. while many of them may claim to be 'gigabit', the truth is they'll never hit such speeds.  I've also seen plenty of instances where switch to switch comms with consumer switches caused slowness issues and dropped packets.  (BTW, it's worth noting that the OP has 185 odd dropped packets, which probably contributed to the original issue.)

 

 

The 185 dropped packets aren't unusual when you're having a cable issue ... and certainly contributes to how slow it was running.

 

The 185 dropped packets aren't unusual when you're having a cable issue ... and certainly contributes to how slow it was running.

That many dropped packets would have absoulutly no noticeable affect on speed. A packet is 1500 bytes or less.

 

Sent from my SGH-I727R using Tapatalk 2

 

 

The 185 dropped packets aren't unusual when you're having a cable issue ... and certainly contributes to how slow it was running.

That many dropped packets would have absoulutly no noticeable affect on speed. A packet is 1500 bytes or less.

 

Sent from my SGH-I727R using Tapatalk 2

 

After switching the cable I've transferred a total of 4TB over to the array and its only reporting 74 packets dropped so I guess that is pretty good.

 

Regarding nics...The one on the server is a built into the dell mobo. I happen to have an Intel GT laying around I could pop in. Would that be preferable to the built in dell nic?

 

Thanks

74 dropped packets out of 4TB transferred is more than acceptable.

 

The mobo built-in NIC is an Intel NIC already, so I don't think adding the Intel GT would be of any benefit.