October 6, 20178 yr So basically I'm currently preclearing a drive to replace a faulty one and I noticed some files missing. All but 3 of my music folders, and one off files here and there. I found a post, see below, where someone ran reiserfsck --rebuild-tree and got their old files back. So my question is this. Is it safe to run that command when there is a faulty disk thats going to be replaced? Or should I wait till preclear and rebuild is done then deal with this? I have backups of a lot of the files but I'm afraid to move them there in case the rebuild brings them back or something else weird happens.
October 6, 20178 yr Community Expert You should start by posting your diagnostics:Tools -> Diagnostics
October 6, 20178 yr Author 34 minutes ago, johnnie.black said: You should start by posting your diagnostics: Tools -> Diagnostics Here they are. I looked deeper into it and it looks like everything on the missing disk is not being emulated even though unraid says so. This freaked me out and i added a free precleared drive i had and started the rebuild parity sync. That was probably rash but I panic about shit I shouldn’t. So right now its rebuilding and the entire content of the drive is not being emulated. Maybe i Shoukd stop the rebuild? diagnostics-20171006-1914.zip
October 7, 20178 yr Community Expert Let the rebuild finish and when it does run xfs_repair on disk1 (md1): https://wiki.lime-technology.com/Check_Disk_Filesystems#Drives_formatted_with_XFS P.S. There are several ATA errors on disks 7, 15 and 20, check cabling but it can also be the Marvell controllers you're using, never seen so many Marvell controllers on the same server, you have 6 total, 5 in use, all disks with ATA errors are on one of them, and they should be avoided with unRAID v6.
October 7, 20178 yr Author 3 hours ago, johnnie.black said: Let the rebuild finish and when it does run xfs_repair on disk1 (md1): https://wiki.lime-technology.com/Check_Disk_Filesystems#Drives_formatted_with_XFS P.S. There are several ATA errors on disks 7, 15 and 20, check cabling but it can also be the Marvell controllers you're using, never seen so many Marvell controllers on the same server, you have 6 total, 5 in use, all disks with ATA errors are on one of them, and they should be avoided with unRAID v6. Yeah I never knew they were a problem until recently. I’ll have to look into alternatives or get a new board. Disk 15 is actually erroring like crazy, pending sectors way up in the thousands and rebuild is saying its going to take a very long time because data is being transferred at 600kb/s. let it ride or would you suggest something else? I only had one parity due to another drive being RMAd so would hate to lose both.
October 7, 20178 yr Community Expert If there are read errors on disk15 during the rebuild, disk1 will have corrupt files, is old disk1 completely dead? If not, did you check SMART?
October 7, 20178 yr Author 37 minutes ago, johnnie.black said: If there are read errors on disk15 during the rebuild, disk1 will have corrupt files, is old disk1 completely dead? If not, did you check SMART? I didn't get the smart data before pulling it but it was basically saying that it was unmountable and had a red X. i moved it around thinking it was a sata or wire issue and still happened. Then i added it to a second computer and tried hard drive sentinel and it said it couldn’t contact the contents or something. Since the data would be corrupt I stopped the rebuild and added the old disk 1 as unassigned drive and it wont mount and its connected to a sata on the board no marvell. So I’m assuming with disk 15 dying that its a wrap? I can probably save disk 15, done it in the past, for dying disks like that by moving the data to another disk.
October 7, 20178 yr Look at a LSI SAS9201-8i. Less than $50 on eBay. Smart errors like pending sectors are not caused by the Marvell controllers. But they can and do corrupt data. Strongly recommend you stop using them before your data is so full of holes as to be useless.
October 7, 20178 yr Author 5 minutes ago, SSD said: Look at a LSI SAS9201-8i. Less than $50 on eBay. Smart errors like pending sectors are not caused by the Marvell controllers. But they can and do corrupt data. Strongly recommend around using them before your data is so full of holes as to be useless. I really have to research this one because i‘ve always seen people mention them but have no idea how they work and thought they were enterprise only so never thought to get them. Funny part is those marvells combined price was probably more than the LSI
October 7, 20178 yr Not sure how many ports you need, but there is a -16i.version (16 ports) that several people on the forums use (me included). Good option if PCIe ports are limited. Needs a PCIe 2.0 x8 or better slot to not limit bandwidth with 16 drives attached (or more accurately, with 16 drives performing simultaneous high speed I/O - like during a parity check).
October 7, 20178 yr Community Expert Post SMART for old disk1 https://wiki.lime-technology.com/Troubleshooting#Obtaining_a_SMART_report
October 7, 20178 yr There is s definite distinction between drive errors, like pending sectors and reallocated sectors, and drive corruption, like errors reported in the errors column of the WebGUI and in the syslog.The former are detected by SMART reports and unaffected by the controller, disk cabling, or other factors - they normally require new drives, but often in the short run, have no affect on your data. The latter can be very damaging to your data, but can be remedied and not require purchasing new drives.
October 7, 20178 yr Author 11 minutes ago, johnnie.black said: Post SMART for old disk1 https://wiki.lime-technology.com/Troubleshooting#Obtaining_a_SMART_report https://pastebin.com/9sNpRseK Quickly heres what I have from hard drive sentinel stored on my pc for the drive. I will read the faq on how to get the smart from an unassigned drive then get that for you.
October 7, 20178 yr Community Expert SMART looks fine, some CRC errors, those are usually from a bad SATA cable. The disk being unmountable only means that there are filesystem issues, not necessarily disk issues. It's possible to get old disk1 in the array and use the new disk to replace disk15 instead, but only if the array data has remain unchanged since you first took it it out, if yes I can post a procedure for you to do.
October 7, 20178 yr Author Just now, johnnie.black said: SMART looks fine, some CRC errors, those are usually from a bad SATA cable. The disk being unmountable only means that there are filesystem issues, not necessarily disk issues. It's possible to get old disk1 in the array and use the new disk to replace disk15 instead, but only if the array data has remain unchanged since you first took it it out, if yes I can post a procedure for you to do. Yea I’ll give it a try. Ive added data since disk one was gone but don’t think the parity has changed.
October 7, 20178 yr Community Expert 1 minute ago, Wavey said: Ive added data since disk one was gone but don’t think the parity has changed If you added data parity changed for sure, in that case best way forward would be to keep old disk1 intact for now, try to copy everything you can from disk15 to other array disks, when done do a new config with old disk1 (with a new SATA cable) and a new disk15, then run a parity sync. Before or after the parity sync you'll still need to run xfs_repair on disk1.
October 7, 20178 yr Author 19 minutes ago, johnnie.black said: If you added data parity changed for sure, in that case best way forward would be to keep old disk1 intact for now, try to copy everything you can from disk15 to other array disks, when done do a new config with old disk1 (with a new SATA cable) and a new disk15, then run a parity sync. Before or after the parity sync you'll still need to run xfs_repair on disk1. In the process of doing this now, thank you for all the help!
October 7, 20178 yr Author 5 hours ago, SSD said: Not sure how many ports you need, but there is a -16i.version (16 ports) that several people on the forums use (me included). Good option if PCIe ports are limited. Needs a PCIe 2.0 x8 or better slot to not limit bandwidth with 16 drives attached (or more accurately, with 16 drives performing simultaneous high speed I/O - like during a parity check). I have 22 drives connected now running on a GA-990FXA-UD3 see specs below i wouldnt mind 24 drive ability in case i wanted to add two more for whatever reason. i have 8 built in so 16 would be just right. Any good tutorials on how to do this and what parts Id need? I am fully in the dark here. Supports AMD AM3+ FX / AM3 Phenom™ II, Athlon™ II series Processors GIGABYTE Ultra Durable™ 4 Technology GIGABYTE Digital Power Engine 8+2 phase CPU VRM power design for AMD high-TDP CPU support GIGABYTE UEFI DualBIOS™ 2-way CrossFire™/ SLI™ Support Realtek ALC889 with High Quality 108dB SNR HD audio Supports Dolby Home Theater audio GIGABYTE On/Off Charge™ for USB devices 4 USB 3.0 ports with GIGABYTE 3X USB power 6 SATA3 and 2 eSATA3 ports with RAID support LAN with high ESD Protection
October 8, 20178 yr You should just need the card and 4 of the SFF-8087 breakout cables. Your motherboard should handle this no problem. I strongly recommend hot-swap style cages like the CSE-M35T-1B's, but that's a whole different topic. Here is a 9201-16i that you might consider. I know nothing of the seller but looks pretty good at first glance. I would recommend reviewing his feedback and making your own decision. http://www.ebay.com/itm/LSI-SAS9201-16i-6Gbps-SAS-SATA-PCI-Express-2-0-x8-Host-Bus-Adapter-Card-/362123934352?epid=1408742130&hash=item545044ba90:g:5RMAAOSwjqVZAmK~ And here is a breakout cables at monoprice. I recommend the shortest that works. 0.5M is preferred, and 0.75M if you need a little longer (if possible). Shorter cables have fewer problems in my experience: https://www.monoprice.com/search/index?keyword=sff-8087 Post back with any questions. Good luck!
Archived
This topic is now archived and is closed to further replies.