Kboogie Posted February 6, 2023 Share Posted February 6, 2023 Having some drive transfer performance issues and at a loss as to how to resolve. Some details on the configuration and observations below, diagnostics attached. Curious to hear what people think. Hardware Lenovo PC, i5-8400T, 8gb ram, 256gb SSD System Cache RocketRAID 2722 SAS card (marvel), sans digital external SAS enclosure, 8 mixed drives, 2 parity Observations Network transfers start at 60mb, reduce to 40mb and then become erratic, reducing to <1mb. Confirmed same behavior with transfers using unraid UI to move data between disks Confirmed same behavior with parity check, starting at 180mb sustained and then after 10 min plummeting down to <1mb When rates drop there is a spike in CPU from resting 10% to 50%. Included some screen shots of what I see bouncing in the top CPU chart during the event. Please help! unraid-diagnostics-20230206-1619.zip Quote Link to comment
JorgeB Posted February 7, 2023 Share Posted February 7, 2023 Both parity disks (and disk4) are SMR, and some units of that specif model are known to have particularly bad performance, if it's a new array you can retest after assigning the other CMR drives only. Quote Link to comment
Kboogie Posted February 7, 2023 Author Share Posted February 7, 2023 Thanks Jorge, appreciate the help. Do you think I need to replace all SMR or just the parity? I would say the performance has always been volatile, but it's never been this bad where things go down to near zero and hang. Quote Link to comment
JorgeB Posted February 7, 2023 Share Posted February 7, 2023 Set a couple of the 8TB WD drives as parity, then compare the transfer speed to other WD and the Seagate drives in the array, problem might be just one of them, some of those perform decently, others are super slow, not sure why. Quote Link to comment
Decto Posted February 7, 2023 Share Posted February 7, 2023 My previous server had a couple of 2.5" SMR drives in the array and these were fine though they were used for fairly static storage with large occasional writes but mostly read only. The parity drive was always CMR though. No issues with parity checks etc. If you have a lot of active writes on a drive while running parity then SMR may become an more of an issue. Is one of you cache drives connected by USB? Has that always been the case? SSK_SSK_Storage_DB98765432116DA-0-0-20230206-1619 cache_256_sata (sdb) /dev/sdb: Unknown USB bridge [0x7825:0xa2a4 (0x1507)] Quote Link to comment
Kboogie Posted February 12, 2023 Author Share Posted February 12, 2023 On 2/7/2023 at 10:37 AM, JorgeB said: Set a couple of the 8TB WD drives as parity, then compare the transfer speed to other WD and the Seagate drives in the array, problem might be just one of them, some of those perform decently, others are super slow, not sure why. Thanks Jorge, After shuffling the drives it does seem to be happier with parity as CMR. In particular, it's better when it comes to disk to disk transfers across members of the SAS controller. I get a sustained 40-46MB/s and it seems stable. However, the USB SATA SSD is still very inconsistent. Any thoughts on this? Is using USB as cache not a best practice? Quote Link to comment
JorgeB Posted February 12, 2023 Share Posted February 12, 2023 We don't recommend USB for array and pool devices, but a decent USB bridge should behave OK for performance, the main USB problem can be disconnects and error handling in general, I've been using a USB SSD as cache in one of my servers for some times without issues, except SMART attributes don't work correctly. Quote Link to comment
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.