January 14, 20179 yr My server is doing its first parity check and I have Fix Common Problems in troubleshooting mode as I had an issue that caused a reboot and this parity check, would it be slowing down the parity check? Its been going for four hours and its barely at 3%
January 14, 20179 yr Author Does anyone have an opinion? Could the Fix Common Problems in troubleshooting mode be slowing my parity check down? Its going so slowly its going to be running for a week or more.
January 14, 20179 yr Author Here you go. I mean I know with this new Supermicro server, all the drives are connected to one HBA, but I didn't expect it to be this slow. tower-diagnostics-20170113-2220.zip
January 14, 20179 yr SAS2LP is only linking @ x4, don't you have a x8 slot? Jan 13 15:23:48 Tower kernel: mvsas 0000:02:00.0: mvsas: PCI-E x4, Bandwidth Usage: 5.0 Gbps That will limit your speed to under 50MB/s, but even in an x8 slot it can't be faster than about 90MB/s due to the 24 disks on a single SAS2LP link.
January 14, 20179 yr Author Yes I have it in a 16x slot actually. Would that be a motherboard BIOS setting then or a setting on the card?
January 14, 20179 yr Author Dont forget, I have one of the SAS2LP connectors connected to the front backplane with 24 disks, the other SAS2LP connector is connected to the rear backplane where there are 6 disks.
January 14, 20179 yr Card links @ the maximum speed available. x16 Mechanically does not mean x16 Electrically, is this your board? Intel® Server Board S5500HCV: Five expansion slots o One PCI Express* Gen 2 slot (x16 Mechanically,x4 Electrically o Two PCI Express* Express* Gen 2 x8 slots o One PCI Express* Gen 1 slot (x8 Mechanically, x4 Electrically) shared with SAS Module slot. This PCI Express* Gen 1 slot is not available when the SAS module slot is in use and vice versa
January 14, 20179 yr Author Yes that is my board, so does that mean I should put it in an 8x slot to get the maximum bandwidth?
January 14, 20179 yr Yes, one of these: o Two PCI Express* Express* Gen 2 x8 slots Not this one o One PCI Express* Gen 1 slot (x8 Mechanically, x4 Electrically)
January 14, 20179 yr Author Got it, thanks. You need to put that 'Buy me a beer' link in your sig like what Squid has cause I'd buy you one right now! Thanks again.
January 14, 20179 yr Author Good thing I ran this parity check, so far its corrected 520 sync errors, wow! And its running much faster now at 52MB/s hopefully it will speed up.
January 14, 20179 yr My worry would be why it had 520 errors that needed correcting. It might just be the unclean shutdown you alluded to but did you see my reply to one of your other threads?
January 15, 20179 yr Author Yes I did see your reply and I am just going to play it by ear for now and see if it is indeed a problem with the SAS2LP card or not. On another point, before, when I had my hardware in a different tower with drive bays, I had two controllers, the SAS2LP and 16 port raid controller that was operating in JBOD mode. So I had the raid controller connected to 16 drives and the SAS2LP connected to 8 and then the remaining six drives running off the motherboard SATA ports. When I ran a parity check it would usually start slow but eventually get up to 80-90MB/s and on my 8TB parity take about a day and a half to complete. Right now, with just the SAS2LP running all my drives, its taken almost 25 hrs to reach 53% so I am wondering if I get another HBA and connect the rear backplane connector to this second HBA, that should speed up the parity check should it not? Or is there a faster HBA I can get? unRAID is saying its going to take just over 19 more hours to finish the parity check, that means going forward my parity checks are going to take over two days to complete.
January 15, 20179 yr Is the current check with the SAS2LP on the x4 slot? It should be considerably faster on the x8 slot, and there are a few things more you can do to improve it, but getting another controller with help a lot if the backplanes (or at least the front one) supports dual linking. Other things to do: Get all the smaller disks on the front backplane so less disks are using that link a soon as possible. Your using the default tunables, usually not optimal especially with so many disks, try these, should be better than current ones (but parity check start will still be limited by the 24 disks on the single expander link, up to the 3TB mark if you re-arrange the bigger disks to the back plane): nr_requests:8 md_num_stripes:4096 md_sync_window:2048 md_sync_thresh:1900
January 15, 20179 yr Author Yes its in the 8x slot now, but its still slower then the combined speed of of the Areca RAID controller I was using before (in a 16x slot) and the same SAS2LP which was also in a 16x slot (probably running at 8x) I had speeds up to and slightly over 100MB/s, at the moment I don't see speeds faster then 60MB/s. If getting a second SAS2LP and connecting just the rear backplane to it will help, then that is what I will do.
January 15, 20179 yr If getting a second SAS2LP and connecting just the rear backplane to it will help, then that is what I will do. It won't help much if the other one stays connected to the front backplane with single link, does the backplane support dual link?
January 16, 20179 yr Author I don't know if it does or not. One of concern that I failed to mention was that when I moved the HBA to the 8x slot, the cable connecting to the rear backplane was kind of stretched because it is just barely long enough and I think this may be related to the problems because I believe the drives that are are being affected are on the rear back plane. I am ordering a new cable off Amazon that should be long enough and until then the server will remain off. What do you suggest I do when I am ready to bring the server back online? New config, rebuild parity? I know all the drive positions so I shouldn't (hopefully lose any data) its a backup server anyway so if I had to blow it away it wouldn't be the end of the world either.
Archived
This topic is now archived and is closed to further replies.