November 19, 20169 yr Hi all So I brought five Seagate Expansion 5TB USB 3.0 drives, I performed some testing and loaded some with data when in the USB3.0 cases and I seemed to get reasonable transfer speeds, of around 100MB/sec so I was happy. I then removed these from my machine and put them into my UNRAID box and pre-cleared them three times, here's an example of the output, they were all pretty similar on each cycle. unRAID Server Pre-Clear of disk /dev/sdb # # Cycle 3 of 3, partition start on sector 64. # # # # # # Step 1 of 5 - Pre-read verification: [9:28:56 @ 146 MB/s] SUCCESS # # Step 2 of 5 - Zeroing the disk: [9:22:13 @ 148 MB/s] SUCCESS # # Step 3 of 5 - Writing unRAID's Preclear signature: SUCCESS # # Step 4 of 5 - Verifying unRAID's Preclear signature: SUCCESS # # Step 5 of 5 - Post-Read verification: [9:29:21 @ 146 MB/s] SUCCESS # # # # # # # # # # # # # # # ############################################################################################################################ # Cycle elapsed time: 28:20:36 | Total elapsed time: 85:01:15 # ############################################################################################################################ ############################################################################################################################ # # # S.M.A.R.T. Status # # # # # # ATTRIBUTE INITIAL CYCLE 1 CYCLE 2 CYCLE 3 STATUS # # 5-Reallocated_Sector_Ct 0 0 0 0 - # # 9-Power_On_Hours 2 30 58 87 Up 85 # # 183-Runtime_Bad_Block 0 0 0 0 - # # 184-End-to-End_Error 0 0 0 0 - # # 187-Reported_Uncorrect 0 0 0 0 - # # 190-Airflow_Temperature_Cel 34 30 29 29 Down 5 # # 197-Current_Pending_Sector 0 0 0 0 - # # 198-Offline_Uncorrectable 0 0 0 0 - # # 199-UDMA_CRC_Error_Count 0 0 0 0 - # # # # # # # ############################################################################################################################ # SMART overall-health self-assessment test result: PASSED # So seems great. So next up I mounted my old drives and started transferring to them and this is the result. Music/Collection/Ozomatli/Street Signs (2004)/03 - Street Signs.flac 27,041,003 100% 8.63MB/s 0:00:02 (xfr#430216, ir-chk=1013/509126) Music/Collection/Ozomatli/Street Signs (2004)/04 - (Who Discovered) America_.flac 33,882,832 100% 7.71MB/s 0:00:04 (xfr#430217, ir-chk=1012/509126) Consistent 8 to 10MB/sec, on really big files they struggle there way up to about 12MB/sec. When running cp (rather than rsync) things seem to be a little bit faster. So I'm really quite confused, I loaded these drives with data much faster when they were in their USB chassis, but now the performance seems to be terrible. I've done some googling and no one seem to be sure if they are SMR, but that seems to be a non issue as they performed a lot better over USB3.0. Maybe I was naive just buying external drives, but I must of done this dozens of times over 10+ years without it ever being an issue. Is there anything I can do?
November 19, 20169 yr Please see Need help? Read me first!, and attach the diagnostics zip. I'm not sure why you feel it must be the fault of the new drives, rather than something wrong with your current hardware.
November 19, 20169 yr Author Please find attached. The rest of the hardware is a few years old and has never been a problem, admittedly this is my first time with UNRAID so maybe there is something I've missed regarding configuration? skynet-diagnostics-20161119-1540.zip
November 19, 20169 yr Author Please see Need help? Read me first!, and attach the diagnostics zip. I'm not sure why you feel it must be the fault of the new drives, rather than something wrong with your current hardware. Well you got me thinking Rob, so I removed the parity drive from the array (I'd suspended parity calculation so I assumed it wasn't impacting anything) now performance is as expected, with 130MB/sec on the larger files. Curious as to why the parity write was bogging everything down so much, the CPU was never being stretched at all.
November 19, 20169 yr Please see Need help? Read me first!, and attach the diagnostics zip. I'm not sure why you feel it must be the fault of the new drives, rather than something wrong with your current hardware. Well you got me thinking Rob, so I removed the parity drive from the array (I'd suspended parity calculation so I assumed it wasn't impacting anything) now performance is as expected, with 130MB/sec on the larger files. Curious as to why the parity write was bogging everything down so much, the CPU was never being stretched at all. Exactly what do you mean when you say you "suspended parity calculation"? unRAID parity is real-time and there is no way to suspend it as long as you have a parity drive assigned. Normal parity writing actually performs 2 reads and 2 writes of both parity and the disk to be written when writing data. There is also a turbo (reconstruct) write which handles this differently at the expense reading all the other disks instead of reading parity and the disk to be written. See Turbo Write. And now that you have removed parity you will have to completely rebuild it.
November 19, 20169 yr I've read some of the discussion about these drives and you're right, nobody seems to know for sure whether or not they use SMR technology but there is a general consensus that they often behave erratically. Since you find them good performers as data discs when you don't have a parity disk assigned (assuming that's what you meant), how about replacing the parity disk with one of a different model or brand?
November 19, 20169 yr I've read some of the discussion about these drives and you're right, nobody seems to know for sure whether or not they use SMR technology but there is a general consensus that they often behave erratically. Since you find them good performers as data discs when you don't have a parity disk assigned (assuming that's what you meant), how about replacing the parity disk with one of a different model or brand? Don't know anything about these specific drives but obviously write performance will improve if you don't have a parity drive.
November 19, 20169 yr Yes indeed. My thinking was that they might be usable as data disks but not appropriate for parity so the cheapest solution might be to replace the one but keep the rest.
November 20, 20169 yr I've found having a fast parity drive is a must. I use SMR 8TB Seagate drives, and 5400rpm 8TB WD Helium drives, but my parity drive is a 7200rpm Seagate NAS drive which is fit for nearly 200MB/s constant.
November 22, 20169 yr Author Please see Need help? Read me first!, and attach the diagnostics zip. I'm not sure why you feel it must be the fault of the new drives, rather than something wrong with your current hardware. Well you got me thinking Rob, so I removed the parity drive from the array (I'd suspended parity calculation so I assumed it wasn't impacting anything) now performance is as expected, with 130MB/sec on the larger files. Curious as to why the parity write was bogging everything down so much, the CPU was never being stretched at all. Exactly what do you mean when you say you "suspended parity calculation"? unRAID parity is real-time and there is no way to suspend it as long as you have a parity drive assigned. Normal parity writing actually performs 2 reads and 2 writes of both parity and the disk to be written when writing data. There is also a turbo (reconstruct) write which handles this differently at the expense reading all the other disks instead of reading parity and the disk to be written. See Turbo Write. And now that you have removed parity you will have to completely rebuild it. "Suspend" was a misinterpretation by me, it was a new array, the system was performing it's first parity sync. While trying to figure out why everything was going so slow I pressed the button to cancel it, at that point I assumed that there was no further parity calculation, as what's the point of performing parity calculation when the drive has never been in sync. Clearly, with hindsight, this was the wrong assumption. Happy to rebuild, I'm pre-clearing another couple of new drives on a 3-cycle routine, I'll be bringing one in as a parity drive at the end of the week, I'll let it sync up and then we'll see if the awful write performance comes back. While I can see there being a smallish hit from parity calcs, on a system that's not I/O or CPU constrained I wouldn't expect it to be all that significant, however my previous experience is with ZFS which I understand is quite different to UNRAID, so I might have more to learn. Thanks everyone.
November 22, 20169 yr While I can see there being a smallish hit from parity calcs, on a system that's not I/O or CPU constrained I wouldn't expect it to be all that significant, however my previous experience is with ZFS which I understand is quite different to UNRAID, so I might have more to learn. ZFS uses RAID5, very different from unRAID, both have their advantages and disadvantages, unRAID doesn't stripe data, so while you can access data in each disk individually (among other advantages) it will never be as fast, search for turbo write as a way to considerably improve write speed at the expense of all disks spinning up for writes.
November 22, 20169 yr Author What model are the new disks? More of the same, I have five in total of these drives. Unfortunately I have no other 5TB or larger drives, so I may have to purchase something else, or look at other options.
November 23, 20169 yr Like I said, since you've already bought them see if they work satisfactorily as data disks but use something else for parity.
Archived
This topic is now archived and is closed to further replies.