Saturday, May 1, 2010

Formatting the large hard drive - continued

As mentioned in the previous post, one of my servers is formatting a 7.5TB RAID-5 volume.

During the first 24 hours, it has completed 13%.
During the next 24 hours, the progress was at 26%.
Next day, it was at 38%, and the day after (last night) it clocked at 50%.
However, this morning, 12 hours later, it was at 75%, and tonight I expect it to be completed.

So in the very last day the system's progress was the same as for the first 4 days. Moreover, I fully expected this to be the case. Can anyone guess why?

6 comments:

Anonymous said...

This is waaay too slow to be usable, don't you think?

DzembuGaijin said...

This sounds like a bit long

Aistis Zen said...

eemm.. does it try to hold the whole parity table in memory and starts to page? :-) head locality over spindle can't be here, right? you have like 11mb/s build speed, which is .. slow anyway.

Aistis Zen said...

well, did it finish?

Unknown said...

Could it be because of the write speed not being even all around the disk? It starts formatting from the innermost rings which are the slowest, and reaches the outermost which are fastest?

Eric Lee Green said...

Just out of curiousity I checked my server logs from when I had a hard drive crash and had to rebuild my 1TB arrays on Linux 2.6.31. The rebuild of the first one started at 00:57:09 and ended at 05:04:25, or roughly 4 hours per terabyte. Note that my system is decent but not exceptional -- it is a Core 2 Duo 2.4Ghz with 8GB of RAM and three 1TB 7200 RPM SATA drives, from whence I've created two 1TB software RAID5 arrays. Interesting how a system that was state-of-the-art three years ago is now merely decent :).

What this implies is that I would need roughly 30 hours to rebuild 7.5TB of RAID-5 arrays with the 2.6.31 kernel. Is this a reasonable assumption, or no? Hmm, time to think about that...