Re: [Moosefs-users] SSDs

[Steve W. <st...@pu...>]

On 03/30/2012 09:45 PM, Quenten Grasso wrote:
> Also as a side note I went though my small test cluster which is 6 machines with 2 disks each (1ru servers) and replaced all of the disks which seemed to have a higher then average fsync than the other disks and this increased my clusters performance considerably and I'm not currently running any raid. I guess this may go without saying however thought I might mention it :)

Is this a common problem... having disks (all the same model from the 
same manufacturer) which exhibit very different fsync and read/write 
performance?  I've had to replace a few disks (they're Hitachi 3TB  
HDS723030ALA640 drives) and the ones I've replaced have fairly dismal 
performance compared to the other disks in the server.  For example, the 
daily average read transfer speeds range from 58MB/s to 72MB/s but on 
the replacement drive it is 29MB/s.  Similarly, write speeds range from 
3.4MB/s to 3.8MB/s (but one outlier is giving 2.8MB/s) and the 
replacement drive is showing 2.3MB/s.  Average fsync times range between 
50,675us and 58,882us while the replacement drive shows 76,516us.

Since the numbers were so far off from the average, I replaced one drive 
two more times with the same results.  Could it just be that different 
batches of the same disk drive have such widely varying 
characteristics?  Or is there something in how MooseFS rebalances that 
could be causing this?  When I replace a drive, I don't mark it for 
removal first; I just leave myself undergoal for a while until all 
necessary chunks are replicated to the replaced drive.

Thanks,
Steve

>
> Quenten Grasso