On Thu, 14 Jun 2012 08:38:27 -0500 Stan Hoeppner <s...@hardwarefreak.com> wrote:
> On 6/14/2012 4:51 AM, Ramon Hofer wrote: > > >> These commands don't match the pastebin. The pastebin shows you > >> creating a 4 disk RAID5 as /dev/md0. > > > > Really :-? > > That kind of (wrong) analysis is one of the many outcomes of severe > lack of sleep, too much to do, and not enough time. ;) Having 3 > response/reply chains going for the same project doesn't help either. > We share fault on that one: you sent 3 emails before I replied to the > first. I replied to all 3 in succession instead of consolidating all > 3 into one response. Normally I'd do that. Here I simply didn't > have the time. So in the future, with me or anyone else, please keep > it to one response/reply. :) Cuts down on the confusion and overlap > of thoughts. Ok I will. I'm still learning the code of conduct for mailing lists ;-) I then start right now. First of all I tried to set the raid5 with the WD 20EARS and didn't have much luck. They led to fail events when mdadm builds the array. They "worked" in my Netgear NV+ with very low r/w rates <5 MB/s (which I now assume is because of the disks. That's why I'm already thinking of buying new disks. I have found these drives at my local dealer (the prices are in Swiss Francs). 2 TB: - Seagate Barracuda 2TB, 7200rpm, 64MB, 2TB, SATA-3 (129.-) - Seagate ST2000DL004/HD204UI, 5400rpm, 32MB, 2TB, SATA-II (129.-) 3 TB: - Seagate Barracuda 3TB, 7200rpm, 64MB, 3TB, SATA-3 (179.-) I think the Seagate Barracuda 3TB are the best value for money and I didn't find any problems that could prevent me from using them as raid drives. Btw. When I tried to set up the WD20EARS mdstat told me that the syncing would take about 6 hours. Hopefully the Barracudas have at least the same rate. Then the process would be finished on maybe less than 9 hours. This seems to be acceptable for my case. > Also, please note that with 2TB drives, the throughput will decrease > dramatically as you fill the disks. If you're copying over 3-4TB of > files, a write rate of 20-30MB/s at the end of the copy process should > be expected, as you're now writing to the far inner tracks, which have > 1/8th or so the diameter of the outer tracks. Aerial density * track > (cylinder) length * spindle RPM = data rate. The aerial density and > RPM are constants. So if I see low rates in the future I can add a new raid5 and get higher throughbput again because the linear raid would write first to the new array? > > Now I only have to setup the details correctly. > > Like the agcount... > > Like I said, it may not make a huge difference, at least when the XFS > is new, fresh. But at it ages (write/delete/write) over time, the > wonky agcount could hurt performance badly. You balked at that > 20MB/s rate which is actually normal. With XFS parms incorrect, a > year from now you could be seeing max 50MB/s and min 5MB/s. Yeah, > ouch. Another reason to set it up properly now :-) > > You really were an incredible help! > > When I'm not such a zombie that I misread stuff, yeah, maybe a little > help. ;) No really. The adventure of enlarging my media server would have ended in total frustration! -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20120615213116.0631a666@hoferr-x61s.hofer.rummelring
