aminorjourney Posted January 22, 2018 Share Posted January 22, 2018 Hi all! So I've got my unRAID 6.3.5 system up and running, and it's configured thusly (short profile below, full profile attached) Model: Custom M/B: Gigabyte Technology Co., Ltd. - Z77X-UP5 TH-CF CPU: Intel® Core™ i7-3770K CPU @ 3.50GHz HVM: Disabled IOMMU: Disabled Cache: 128 kB, 1024 kB, 8192 kB Memory: 32 GB (max. installable capacity 32 GB) Network: eth0: 1000 Mb/s, full duplex, mtu 1500 eth1: 10000 Mb/s, full duplex, mtu 1500 Kernel: Linux 4.9.30-unRAID x86_64 OpenSSL: 1.0.2k I've currently got two NICs as you can see. The first is set up for general network access, and is connected to the outside world through the standard gigabit switch I have in my office. The second NIC (10GBe) is configured with a static IP and is connected directly to my new iMac Pro, which has its own static address on the same network. This makes it possible for me to transfer files directly between the two computers at speeds in excess of 250 MB/s I've tried using AFS, SMB, and NFS, but as soon as I push the network hard, the shares disappear. In the case of SMB and AFS shares, it causes the computer to crash. NFS doesn't cause a crash, but it does make things a little unhappy. If I turn off my cache drive (512 GB NVMe drive attached on PCI-e bus with a 4cx adaptor) then the shares aren't ejected or suffer any issues, presumably because I'm not pushing the NIC too hard. If I turn the cache drive On (or to prefer or only, for example) then the system gets all upset, eventually causing the share to eject. I'm making this post because I suspect the occam's razor answer here is that I'm using a bad ethernet cable, and that there's significant packet loss at high transfer speeds. BUT I want to make sure I'm not missing anything else. I'm not in the office for a few days, so it's going to take a while to get an answer. But if anyone has any suggestions/ideas/experience I haven't got, please share. Thanks in advance! Nikki. Profile.xml Link to comment
Vr2Io Posted January 22, 2018 Share Posted January 22, 2018 11 minutes ago, aminorjourney said: I suspect the occam's razor answer here is that I'm using a bad ethernet cable, and that there's significant packet loss at high transfer speeds You may further check does lot of pkt error record by unRAID in "Dashboard" -> "Network" -> "System status (pull down - error)" I have similar setup, but i5-3570K was a PC. If 1 session file transfer (SMB) also ~300MB, the overall speed will increase with more session transfer but PC will stuck / lag. BTW, my unRAID NVMe / system never got share eject or crash. ** remarks : iperf test at both end can reach 900MB+ and no lag ** Link to comment
aminorjourney Posted January 22, 2018 Author Share Posted January 22, 2018 I can't see that particular setting -- I can see the Network and system status... but that's it. ...and I think that doesn't work for me since I just reset the server, and it's since last reboot... so I'll run some more tests and see if those errors appear. I've turned the cache disk off to see if that helps things a little. I'm also wondering if the MTU setting needs changing on eth1. I did a ping test earlier and got no packet loss with ping -D -s 1472 <server> so ... I don't *think* it's the issue. Like I said, I'm going to get a nice Cat 6e later, as I suspect my cabling isn't up to par. I've got 10 days left to sort this out, so fingers crossed I can do it before I decide to buy this Link to comment
JorgeB Posted January 22, 2018 Share Posted January 22, 2018 3 minutes ago, aminorjourney said: I'm also wondering if the MTU setting needs changing on eth1 If all the hardware used supports it jumbo frames are recommended for 10GbE. Link to comment
aminorjourney Posted January 22, 2018 Author Share Posted January 22, 2018 Yes. That's what I thought. I did try an MTU of 9,000 last week (so jumbo frame) and had the same fallover. Which brings me back to suspecting an issue with the actual cable. I have to work from home today as I'm looking after a kid, but i'm going to try and drop by the local cable store later and see if I can get a brand-new cat 6e I can use. Link to comment
aminorjourney Posted January 22, 2018 Author Share Posted January 22, 2018 Yes. That's what I thought. I did try an MTU of 9,000 last week (so jumbo frame) and had the same fallover. Which brings me back to suspecting an issue with the actual cable. I have to work from home today as I'm looking after a kid, but i'm going to try and drop by the local cable store later and see if I can get a brand-new cat 6e I can use. Update: Been watching the server remotely via our VPN... and interestingly the built-in Gigabit Nic is having issues, not the 10GBe. Link to comment
Vr2Io Posted January 22, 2018 Share Posted January 22, 2018 Anyway no error found. So, I don't think problem on ethernet cable. What model of NVMe ? And file system was BTRFS ? As you have 32G RAM, you cay try cache a big file to make some more test and better identify the problem. Link to comment
aminorjourney Posted January 22, 2018 Author Share Posted January 22, 2018 The NVMe is a 512 GB Samsung Pro 960. It's formatted with BTRFS. And yeah, I may try that with the RAM -- look up a way to make that as my linux skills are a little rusty! I should also add here that I've been watching the network connection, and it looks like things are okay for now. I will wait for the current work to finish (transcoding large video files, slowly) and then I'll tweak the network settings for jumbo packets. Link to comment
Vr2Io Posted January 22, 2018 Share Posted January 22, 2018 Have you try "scrub" check on NVMe ( click cache and found scrub ), just want to make loading to see share eject or crash. Link to comment
Recommended Posts
Archived
This topic is now archived and is closed to further replies.