On 12/18/2010 12:00 AM, Lamar Owen wrote: > On Friday, December 17, 2010 06:14:31 pm Terry Barnaby wrote: >> The two main RAID1 disks are WD10EARS (Green). I have seen reported some >> issues with the performance of these but in my case they appear to work >> fine when the system is running ok. > [snip] >> Anyone seen this sort of behaviour before ? >> Any ideas one where to look ? > > Yes, I have. > > Use a different drive. Use iostat -x 1 to trace which disk in the RAID1 is > causing problems; you'll likely find that the WD10EARS are throwing long > awaits. Rumor is that this is by design; WD has enterprise 'RAID ready' > drives and don't rate the lower priced drives for RAID. I have a WD15EADS > that does this. At least the EARS version can possibly be put in a 'TLER' > mode that allows RAID use. > > In my case, I had the WD15EADS drive as one half of a RAID1, with the other > half being a Seagate 1.5TB drive of the same LBA. Every once in a while, > performance would absolutely go to pot, and stay that way for minutes at a > time (load averages>10 on a single core system). Using iostat -x 1 I was > able to isolate the issue to that particular drive (I swapped controller > channels, swapped cables, swapped out the power supply, swapped to a > different controller chip on the motherboard, swapped motherboards, and the > issue was always on this drive). > > When I replaced the WD15EADS with another Seagate 1.5TB, performance came > back to normal. I'm using the WD15EADS in a single mode, now, with much > lighter usage, and realizing that performance is not its strong suite. > > Also, the EARS version might use 4K sectors, exposing 512 byte sectors in an > 'emulation' mode; properly aligning partitions to 4K boundaries solves that. > > Google 'WD EARS TLER' and get the whole story. You'll also want to disable > the 'green' mode, as that will also negatively impact performance. There are > tools out there to do that.
Thanks for the info. I did play with setting a partition on a 4096 byte (8 x 512Byte sector) boundary, but saw no change in random 512Byte block write speed with a simple test program. These are recent drives so I wondered if things had changed in this regard. It is strange, however, how the system can run perfectly fine with good fast disk IO for a while and then go into this slow mode. In the slow mode a command can take 30seconds or more to run on an unloaded system. It smacks of some Linux kernel SATA driver/RAID1 versus WD EARS drive interaction to me. However, I think I will change the drives. I was hoping to try some WD10EADS ones I have, but after your issues I will look at the RE series or another make ... Cheers Terry -- users mailing list users@lists.fedoraproject.org To unsubscribe or change subscription options: https://admin.fedoraproject.org/mailman/listinfo/users Guidelines: http://fedoraproject.org/wiki/Mailing_list_guidelines
