kakashisensei Posted July 15, 2020 Posted July 15, 2020 (edited) I currently am using an 6 port Asmedia 1062+1093*2 sata controller for extra ports. Extra ports are for unassigned devices and 1 or 2 array drives. It runs off a mini pcie 2.0 x1 port via a mpcie to pcie riser. The 1062 is a pcie x2 chipset for 2 sata ports, and the 1093 chipsets are port multipliers. The cache and majority of array drives are using onboard 4 port intel sata controller. My motherboard is an old mini itx sandy/ivybridge era, only has one pcie 16x slot and that mini pcie x1 slot. Need to run the gpu off the pcie 16x slot. The Asmedia controller was working fine for ~6months. Recently one of my array drives died, an 8 yr old WD green 2tb. It was getting many exception emask frozen and failed command: WRITE FPDMA QUEUED errors in the log. This was connected to the Asmedia controller and the drive would often get dropped after alot of errors. Unraid would not detect the drive until a reboot. And sometimes, the other drives on the Asmedia could be dropped as well even though they are fine. Luckily those were unassigned devices and not array drives. Sometimes that drive would seem like its fine for a while, and then spit alot of these errors again. I suspect the drive dropping is due to the Asmedia controller. Looking for a cheap and reliable sata controller and it seems the JMB585 cards have been reported to work fine. I already have a dell h310 and ibm m1015, but they simply refuse to work off that mpcie x1 port. All the other non lsi hba cards I've tried work off that x1 port. I've tried everything I can think of but nothing works. Would the JMB585 be better and more stable than the ASM1062? Edited July 15, 2020 by kakashisensei Quote
Vr2Io Posted July 15, 2020 Posted July 15, 2020 (edited) JMB585 should replace Asmesia 1062+1093 solution. In general, port multipliers, Marvell 92xx, USB bridge ... are not recommend for Unraid. If you can't well test its reliability then you should avoid them. For about disk, you need keep close monitor SMART, if any abnormal sign occur, you should replace it. I usually change some age disk be backup disk periodic. I also health check all disks in monthly. When first meet Unraid, I use Marvell 9215 which immediately cause trouble, then I start use LSI HBA until now. But I also have another build use Marvell 9215+USB bridge which totally form a 13 disk system that haven't give me trouble in yrs+. Recently, I buy a JMB585 M2, due to price drop but I haven't try or test it, just keep as spare. Edited July 15, 2020 by Benson Quote
kakashisensei Posted July 15, 2020 Author Posted July 15, 2020 What are some typical problems that occur with port multipliers? Quote
JorgeB Posted July 15, 2020 Posted July 15, 2020 Frequent ATA errors on the log, which result in freezes and can even drop devices in some cases, Asmedia is good as a 2 port controller, not good with port multipliers, for more ports get a JMB585 or an 8 port LSI HBA. Quote
kakashisensei Posted July 23, 2020 Author Posted July 23, 2020 (edited) I got one of these pcie switch cards and now can run both the gpu and the lsi hba off the single pcie slot. Asmedia pcie switch card I am using this with m.2 to pcie powered riser extensions. Had to use acs override to break up just that one cpu pcie root port, but it works great. I can pass the gpu to VM while the hba card stays on host. For graphics benchmarks that arent bandwidth intensive, i see roughly 1-2% less performance due to the added latency having to go through the pcie switch. My gpu was already running x4 link only, as I have it externally mounted in a separate case with its own power supply. The lower bandwidth doesnt affect the performance for what I use it for. Probably not worth the cost, but I plan to reuse this switch card down the road when I move my desktop parts to upgrade the unraid server. That motherboard is also mini itx, but it supports 2 way split x8 x8 bifurcation. I could run 3 pcie devices off one pcie x16 slot! Might be handy in a couple years when 5/10Gbps is available and present in cheaper switches. On another note, I started seeing these errors on one of my unassigned devices when connected to the lsi hba. Did not see this before when its connected to the intel ich or the asm1062. This drive does have huge and always increasing dma crc error count that I could never resolve it. Tried different cables and cleaning the connectors, but it didn't resolve it. But this drive has always passed extended smart tests. Lsi hba is flashed to IT mode with the latest P20 firmware. I will try switching the controllers to see if it goes away. Also bought the JMB585 card to test. kernel: sd 7:0:0:0: [sdf] tag#1505 UNKNOWN(0x2003) Result: hostbyte=0x00 driverbyte=0x00 kernel: sd 7:0:0:0: [sdf] tag#1505 CDB: opcode=0x28 28 00 61 b2 a6 98 00 00 88 00 kernel: print_req_error: I/O error, dev sdf, sector 1639098008 Edited July 23, 2020 by kakashisensei Quote
Vr2Io Posted July 23, 2020 Posted July 23, 2020 Does error only happen when disk change from spindown to spiinup ? Quote
kakashisensei Posted July 23, 2020 Author Posted July 23, 2020 (edited) It might be, but disk spin down/up is erratic due to it being used on an VM. These errors appear typically appear after a few hrs from each other, but no consistent interval. I'll try to set it up so it spins up when I can monitor it. Edited July 23, 2020 by kakashisensei Quote
kakashisensei Posted July 26, 2020 Author Posted July 26, 2020 (edited) Those errors with the UD disk and the lsi sas2008 don't occur when its spin up/down. Can't figure exactly a pattern to when they occur. I put that drive on the intel ich or jmb585 and doesn't show that error any more. Disk is a toshiba 2.5" laptop hd. It has a weird behavior that unraid shows it spun down / no green light, but it hasn't spun down. The drive doesn't follow the spin down delay I set. On another note, I am getting errors with my new hitachi hc320 with the jmb585 only. They are exception emask/ READ FPDMA QUEUED / hard resetting link. This was only with testing the jmb585 and diskspeed. The benchmark runs fine and seems to be no problem, other than the errors showing up in log. The CRC error count also went up from 0. Tried different port/cables but that didn't resolve it. Haven't used the jmb585 for further testing. No other drives show errors with the jmb. I'll just stick to the lsi card now that its working with a pcie switch, and use ssd and that UD disk off the intel ich. Edited July 26, 2020 by kakashisensei Quote
Vr2Io Posted July 26, 2020 Posted July 26, 2020 I have similar experience with problem 9215, only WD disk will got CRC error / hard resetting link ( never disk drop ), it haven't problem with onboard / Asmedia 1061. After change to LSI, I haven't problem with different brand disk ( HGST / Toshiba / WD ) even change platform several times. 30 minutes ago, kakashisensei said: I'll just stick to the lsi card now Sound good, I stick for longtime. Quote
JorgeB Posted July 26, 2020 Posted July 26, 2020 There can always be some disk/controller combinations that don't work correctly, e.g., I had issues before with some HGST 2.5" drives and LSI, 2.5" Seagate SMR drives and Asmedia, and another model with JMB, but forget the model. Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.